Open-source Grok Bot alternative for named AI teammates. Local-first desktop, web, and mobile clients with per-bot tools, instructions, and memory. Independent fork of T3 Code.
feat: honor model-specific reasoning on every provider and remove the token hard stop (#353)
Leob2ba9a88add2
diff --git a/.changeset/icy-impalas-talk.md b/.changeset/icy-impalas-talk.mdnew file mode 100644index 000000000..4c517ea4f--- /dev/null+++ b/.changeset/icy-impalas-talk.md@@ -0,0 +1,5 @@+---+"akeru-bot": patch+---++Fix model-specific reasoning controls and runtime forwarding across all providers, with saved bot choices shared by web, desktop, and mobile.diff --git a/.changeset/quiet-otters-rest.md b/.changeset/quiet-otters-rest.mdnew file mode 100644index 000000000..87fa84a58--- /dev/null+++ b/.changeset/quiet-otters-rest.md@@ -0,0 +1,5 @@+---+"akeru-bot": patch+---++Remove the per-bot token hard stop. Bots no longer refuse new messages after a set token count, and bot settings and usage on web, desktop, and mobile no longer show a cap.diff --git a/apps/mobile/src/features/threads/ThreadComposer.tsx b/apps/mobile/src/features/threads/ThreadComposer.tsxindex df97585a5..658a01f77 100644--- a/apps/mobile/src/features/threads/ThreadComposer.tsx+++ b/apps/mobile/src/features/threads/ThreadComposer.tsx@@ -1,4 +1,3 @@-import { Predicate } from "effect";import { useMobileI18n } from "../../lib/i18n";import type {EnvironmentId,@@ -42,29 +41,23 @@ import {import { ControlPill } from "../../components/ControlPill";import { ProviderIcon } from "../../components/ProviderIcon";import type { DraftComposerImageAttachment } from "../../lib/composerImages";-import { buildModelOptions, groupByProvider, resolveModelSendBlock } from "../../lib/modelOptions";+import { resolveModelSendBlock } from "../../lib/modelOptions";import { useScaledTextRole } from "../settings/appearance/useScaledTextRole";import { useAppearancePreferences } from "../settings/appearance/AppearancePreferencesProvider";import type { RemoteClientConnectionState } from "../../lib/connection";-import { resolveProviderOptionDescriptors } from "../../lib/providerOptions";import { useComposerPathSearch } from "../../state/use-composer-path-search";-import { botEnvironment, environmentBotsAtom, environmentGroupsAtom } from "../../state/bots";+import { environmentBotsAtom, environmentGroupsAtom } from "../../state/bots";import { providerBotName } from "./thread-list-v2-items";import { resolveThreadIdentity } from "./threadIdentity";import { serverEnvironment } from "../../state/server";import { useEnvironmentQuery } from "../../state/query";-import { useAtomCommand } from "../../state/use-atom-command";-import { squashAtomCommandFailure } from "@akeru/client-runtime/state/runtime";import { ComposerCommandPopover, type ComposerCommandItem } from "./ComposerCommandPopover";import { buildComposerCommandItems } from "./composer-command-items";import { composerMentionItemToken, isThreadMentionQuery } from "./composerMentionItems";import { ComposerMentionPopover } from "./ComposerMentionPopover";-import {- type ExistingThreadSettingsRouteSession,- useExistingThreadSettingsRoutePresentation,-} from "./ThreadSettingsSheet";+import { useExistingThreadSettingsRoutePresentation } from "./ThreadSettingsSheet";import { useThreadSettingsSheetRoute } from "./use-thread-settings-sheet-presentation";-import { buildBotUsageCapPatch } from "./botStepUsage";+import { useThreadComposerSettings } from "./use-thread-composer-settings";import {ComposerConnectionStatusPill,composerConnectionStatus,@@ -175,8 +168,6 @@ export const ThreadComposer = memo(function ThreadComposer(props: ThreadComposer? composerIdentity.title: (bot?.name ?? providerBotName(composerProviderDriver));- const updateBot = useAtomCommand(botEnvironment.update, { reportFailure: false });- const deleteBot = useAtomCommand(botEnvironment.delete, { reportFailure: false });const wasExpandedBeforePreviewRef = useRef(false);const inFlightThreadIdsRef = useRef(new Set<string>());const { onExpandedChange } = props;@@ -248,9 +239,6 @@ export const ThreadComposer = memo(function ThreadComposer(props: ThreadComposerconst sendLabel =props.connectionState !== "connected" || props.queueCount > 0 ? "Queue" : "Send";- const currentModelSelection = props.selectedThread.modelSelection;- const currentRuntimeMode = props.selectedThread.runtimeMode;-const connectionStatus = composerConnectionStatus({connectionError: props.connectionError,connectionState: props.connectionState,@@ -413,123 +401,18 @@ export const ThreadComposer = memo(function ThreadComposer(props: ThreadComposer[composerTrigger, draftMessage, onChangeDraftMessage],);- // ── Model menu ───────────────────────────────────────────- const modelOptions = useMemo(- () => buildModelOptions(props.serverConfig, currentModelSelection, subscriptionStatuses, t),- [props.serverConfig, currentModelSelection, subscriptionStatuses, t],- );-- const providerGroups = useMemo(() => groupByProvider(modelOptions), [modelOptions]);-- // An existing thread is bound to its harness: sessions can't move between- // provider instances, so the picker only offers the thread's own group.- const threadProviderGroups = useMemo(- () => providerGroups.filter((group) => group.providerKey === currentModelSelection.instanceId),- [providerGroups, currentModelSelection.instanceId],- );-- const currentModelOption =- modelOptions.find(- (option) =>- option.selection.instanceId === currentModelSelection.instanceId &&- option.selection.model === currentModelSelection.model,- ) ?? null;-- const providerOptionDescriptors = useMemo(- () =>- resolveProviderOptionDescriptors({- capabilities: currentModelOption?.capabilities,- selections: currentModelSelection.options,- }),- [currentModelOption?.capabilities, currentModelSelection.options],- );-- const updateBotUsageCap = useCallback(- async (input: string) => {- if (!bot) return false;- const patch = buildBotUsageCapPatch(bot.id, input, currentModelOption?.providerDriver);-- if (!patch) return false;- const result = await updateBot({ environmentId: props.environmentId, input: patch });-- return Predicate.isTagged(result, "Success");- },- [bot, currentModelOption?.providerDriver, props.environmentId, updateBot],- );-- const deleteThreadBot = useCallback(async () => {- if (!bot) return t("The command failed.");-- const result = await deleteBot({- environmentId: props.environmentId,- input: { botId: bot.id },- });-- if (!Predicate.isTagged(result, "Failure")) return null;- const error = squashAtomCommandFailure(result);-- return error instanceof Error ? error.message : t("The command failed.");- }, [bot, deleteBot, props.environmentId, t]);-- const settingsRouteSession = useMemo<ExistingThreadSettingsRouteSession>(- () => ({- ownerId: settingsOwnerId,- providerGroups: threadProviderGroups,- selectedModel: currentModelSelection,- onSelectModel: (option) => props.onUpdateModelSelection(option.selection),- optionDescriptors: providerOptionDescriptors,- onUpdateOptionSelections: (options) =>- props.onUpdateModelSelection({ ...currentModelSelection, options }),- runtimeMode: currentRuntimeMode,- onUpdateRuntimeMode: props.onUpdateRuntimeMode,- ...(bot- ? {- memoryThreadRef: {- environmentId: props.environmentId,- threadId: props.selectedThread.id,- },- routinesRef: {- environmentId: props.environmentId,- botId: bot.id,- botName: bot.name,- },- }- : {}),- ...(bot- ? {- botUsageCap: bot.usageCap,- botUsageCapProviderDriver: currentModelOption?.providerDriver,- onUpdateBotUsageCap: updateBotUsageCap,- onDeleteBot: deleteThreadBot,- }- : {}),- }),- [- currentModelSelection,- currentRuntimeMode,- bot,- deleteThreadBot,- props.environmentId,- props.onUpdateModelSelection,- props.onUpdateRuntimeMode,- props.selectedThread.id,- providerOptionDescriptors,- settingsOwnerId,- threadProviderGroups,- updateBotUsageCap,- ],- );-- const openSettings = useCallback(() => {- settingsRoutePresentation.present(settingsRouteSession);- settingsSheetPresentation.open();- }, [settingsRoutePresentation.present, settingsRouteSession, settingsSheetPresentation.open]);-- useEffect(() => {- if (settingsSheetPresentation.isActive) {- settingsRoutePresentation.present(settingsRouteSession);- }- }, [settingsRoutePresentation.present, settingsRouteSession, settingsSheetPresentation.isActive]);+ const { currentModelSelection, currentModelOption, openSettings } = useThreadComposerSettings({+ environmentId: props.environmentId,+ selectedThread: props.selectedThread,+ serverConfig: props.serverConfig,+ subscriptionStatuses,+ bot,+ settingsOwnerId,+ settingsRoutePresentation,+ settingsSheetPresentation,+ onUpdateModelSelection: props.onUpdateModelSelection,+ onUpdateRuntimeMode: props.onUpdateRuntimeMode,+ });return (<Animated.Viewdiff --git a/apps/mobile/src/features/threads/ThreadSettingsSheet.tsx b/apps/mobile/src/features/threads/ThreadSettingsSheet.tsxindex d4aaafb14..0f76fca65 100644--- a/apps/mobile/src/features/threads/ThreadSettingsSheet.tsx+++ b/apps/mobile/src/features/threads/ThreadSettingsSheet.tsx@@ -2,6 +2,13 @@ import { useNavigation } from "@react-navigation/native";import type { NativeStackNavigationProp } from "@react-navigation/native-stack";import { useEffect, useMemo } from "react";import { View } from "react-native";+import { botEngineModelSelection } from "@akeru/shared/model";+import { useMobileI18n } from "../../lib/i18n";+import { buildModelOptions } from "../../lib/modelOptions";+import { useThreadEngineBot } from "../../state/use-thread-engine-bot";+import { useRemoteEnvironmentRuntime } from "../../state/use-remote-environment-registry";+import { useEnvironmentQuery } from "../../state/query";+import { serverEnvironment } from "../../state/server";import { resolveProviderOptionDescriptors } from "../../lib/providerOptions";import { useNewTaskFlow } from "./new-task-flow-provider";import {@@ -9,6 +16,7 @@ import {useExistingThreadSettingsRoutePresentation,} from "./thread-settings-session";import { ThreadSettingsPickerNavigator } from "./thread-settings-picker";+import { liveBotThreadSettings } from "./thread-settings-bot-engine";export {ExistingThreadSettingsRouteProvider,@@ -22,6 +30,26 @@ export function ExistingThreadSettingsRouteScreen() {const navigation = useNavigation<NativeStackNavigationProp<{ ThreadSettings: undefined }>>();const presentation = useExistingThreadSettingsRoutePresentation();const session = presentation.session;+ const { t } = useMobileI18n();+ const engineBotRef = session?.engineBotRef;+ const bot = useThreadEngineBot(engineBotRef?.environmentId, engineBotRef);+ const runtime = useRemoteEnvironmentRuntime(engineBotRef?.environmentId ?? null);+ const selection = bot?.engine ? botEngineModelSelection(bot.engine) : session?.selectedModel;++ // Same account check as the composer, so the sheet never offers a model Send would block.+ const subscriptionAuth = useEnvironmentQuery(+ engineBotRef+ ? serverEnvironment.subscriptionAuth({ environmentId: engineBotRef.environmentId, input: {} })+ : null,+ );++ const subscriptionStatuses = subscriptionAuth.data?.providers;++ const liveModels = useMemo(+ () =>+ buildModelOptions(runtime?.serverConfig ?? null, selection ?? null, subscriptionStatuses, t),+ [runtime?.serverConfig, selection, subscriptionStatuses, t],+ );useEffect(() => {if (session) {@@ -35,7 +63,11 @@ export function ExistingThreadSettingsRouteScreen() {return <View className="flex-1 bg-sheet" />;}- const { ownerId: _ownerId, ...settings } = session;+ const {+ ownerId: _ownerId,+ engineBotRef: _engineBotRef,+ ...settings+ } = liveBotThreadSettings(session, bot, liveModels);return (<ThreadSettingsSessionProvider {...settings}>diff --git a/apps/mobile/src/features/threads/botStepUsage.test.ts b/apps/mobile/src/features/threads/botStepUsage.test.tsdeleted file mode 100644index 22cfdced0..000000000--- a/apps/mobile/src/features/threads/botStepUsage.test.ts+++ /dev/null@@ -1,35 +0,0 @@-import { BotId } from "@akeru/contracts";-import { describe, expect, it } from "vite-plus/test";--import { buildBotUsageCapPatch, resolveBotUsageCapForProvider } from "./botStepUsage";--describe("mobile bot step usage", () => {- it("builds set and clear patches without changing other bot fields", () => {- const botId = BotId.make("bot-1");-- expect(buildBotUsageCapPatch(botId, "50000")).toEqual({- botId,- usageCap: { unit: "tokens", limit: 50_000 },- });- expect(buildBotUsageCapPatch(botId, " ")).toEqual({ botId, usageCap: null });- expect(buildBotUsageCapPatch(botId, "1.5")).toBeNull();- expect(buildBotUsageCapPatch(botId, "0")).toBeNull();- });-- it("clears hard stops for occupancy-only providers", () => {- const botId = BotId.make("bot-1");-- expect(resolveBotUsageCapForProvider("50000", "grok")).toEqual({- available: false,- limit: null,- });- expect(buildBotUsageCapPatch(botId, "50000", "grok")).toEqual({- botId,- usageCap: null,- });- expect(buildBotUsageCapPatch(botId, "50000", "codex")).toEqual({- botId,- usageCap: { unit: "tokens", limit: 50_000 },- });- });-});diff --git a/apps/mobile/src/features/threads/botStepUsage.ts b/apps/mobile/src/features/threads/botStepUsage.tsdeleted file mode 100644index 33bac4f29..000000000--- a/apps/mobile/src/features/threads/botStepUsage.ts+++ /dev/null@@ -1,37 +0,0 @@-import type { UpdateBotInput } from "@akeru/client-runtime/state/bots";-import type { BotId } from "@akeru/contracts";--type ProviderUsageCap = {- readonly available: boolean;- readonly limit: number | null | undefined;-};--export function parseBotUsageCapInput(input: string): number | null | undefined {- if (input.trim().length === 0) return null;- const limit = Number(input);-- return Number.isSafeInteger(limit) && limit > 0 ? limit : undefined;-}--export function resolveBotUsageCapForProvider(- input: string,- providerDriver?: string,-): ProviderUsageCap {- if (providerDriver === "grok") {- return { available: false, limit: null };- }-- return { available: true, limit: parseBotUsageCapInput(input) };-}--export function buildBotUsageCapPatch(- botId: BotId,- input: string,- providerDriver?: string,-): UpdateBotInput | null {- const { limit } = resolveBotUsageCapForProvider(input, providerDriver);-- if (limit === undefined) return null;-- return { botId, usageCap: limit === null ? null : { unit: "tokens", limit } };-}diff --git a/apps/mobile/src/features/threads/thread-feed-rows.tsx b/apps/mobile/src/features/threads/thread-feed-rows.tsxindex 456162e56..d08be7764 100644--- a/apps/mobile/src/features/threads/thread-feed-rows.tsx+++ b/apps/mobile/src/features/threads/thread-feed-rows.tsx@@ -368,9 +368,7 @@ export function renderFeedEntry(function BotStepMeter(props: { readonly meter: BotStepMeterData }) {const label = `${formatBotStepEngine(props.meter.engine)} · ${props.meter.tokens === null ? "—" : formatTokens(props.meter.tokens)- } tokens · ${props.meter.costUsd === null ? "$—" : formatUsd(props.meter.costUsd)}${- props.meter.hardStopReached ? " · Hard stop" : ""- }`;+ } tokens · ${props.meter.costUsd === null ? "$—" : formatUsd(props.meter.costUsd)}`;return (<Textdiff --git a/apps/mobile/src/features/threads/thread-settings-bot-engine.test.ts b/apps/mobile/src/features/threads/thread-settings-bot-engine.test.tsnew file mode 100644index 000000000..738796c25--- /dev/null+++ b/apps/mobile/src/features/threads/thread-settings-bot-engine.test.ts@@ -0,0 +1,87 @@+import { EnvironmentId, ProviderInstanceId } from "@akeru/contracts";+import { describe, expect, it, vi } from "vite-plus/test";+import { makeMobileBot } from "../../lib/mobile-fixtures.test-support";+import type { ModelOption } from "../../lib/modelOptions";+import type { ExistingThreadSettingsRouteSession } from "./thread-settings-session";+import { liveBotThreadSettings } from "./thread-settings-bot-engine";++const model: ModelOption = {+ key: "codex:new-model",+ label: "New model",+ subtitle: "Codex",+ providerKey: "codex",+ providerLabel: "Codex",+ providerDriver: "codex",+ isDefault: false,+ isLegacy: false,+ disabledReason: null,+ selection: { instanceId: ProviderInstanceId.make("codex"), model: "new-model" },+ capabilities: {+ optionDescriptors: [+ {+ id: "reasoningEffort",+ label: "Reasoning",+ type: "select",+ currentValue: "default",+ options: [+ { id: "default", label: "Provider default", isDefault: true },+ { id: "low", label: "Low" },+ { id: "high", label: "High" },+ ],+ },+ ],+ },+};++function settings(): ExistingThreadSettingsRouteSession {+ return {+ ownerId: "environment:chat",+ engineBotRef: { environmentId: EnvironmentId.make("environment"), botId: "bot-1" },+ providerGroups: [],+ selectedModel: { instanceId: ProviderInstanceId.make("codex"), model: "old-model" },+ onSelectModel: vi.fn(),+ optionDescriptors: [],+ onUpdateOptionSelections: vi.fn(),+ runtimeMode: "full-access",+ onUpdateRuntimeMode: vi.fn(),+ };+}++describe("live bot settings sheet", () => {+ it("shows a remote change and reset independently of the frozen composer snapshot", () => {+ const session = settings();++ const bot = makeMobileBot({+ engine: {+ provider: "codex",+ model: "new-model",+ options: [{ id: "reasoningEffort", value: "high" }],+ },+ });++ const live = liveBotThreadSettings(session, bot, [model]);++ expect(live.selectedModel?.model).toBe("new-model");+ expect(live.optionDescriptors[0]?.currentValue).toBe("high");+ live.onUpdateOptionSelections([{ id: "reasoningEffort", value: "low" }]);+ expect(session.onSelectModel).toHaveBeenCalledWith({+ ...model,+ selection: { ...model.selection, options: [{ id: "reasoningEffort", value: "low" }] },+ });++ const reset = liveBotThreadSettings(+ session,+ { ...bot, engine: { provider: "codex", model: "new-model" } },+ [model],+ );++ expect(reset.optionDescriptors[0]?.currentValue).toBe("default");+ });++ it("leaves group and plain settings unchanged without a direct engine bot reference", () => {+ const { engineBotRef: _reference, ...session } = settings();+ const bot = makeMobileBot({ engine: { provider: "codex", model: "new-model" } });++ expect(liveBotThreadSettings(session, bot, [model])).toBe(session);+ });+});diff --git a/apps/mobile/src/features/threads/thread-settings-bot-engine.ts b/apps/mobile/src/features/threads/thread-settings-bot-engine.tsnew file mode 100644index 000000000..e1f2c88f2--- /dev/null+++ b/apps/mobile/src/features/threads/thread-settings-bot-engine.ts@@ -0,0 +1,38 @@+import type { OrchestrationBot } from "@akeru/contracts";+import { botEngineModelSelection } from "@akeru/shared/model";+import { groupByProvider, type ModelOption } from "../../lib/modelOptions";+import { resolveProviderOptionDescriptors } from "../../lib/providerOptions";+import type { ExistingThreadSettingsRouteSession } from "./thread-settings-session";++/** The visible sheet follows bot snapshots even while its composer screen is frozen. */+export function liveBotThreadSettings(+ session: ExistingThreadSettingsRouteSession,+ bot: OrchestrationBot | null,+ models: ReadonlyArray<ModelOption>,+): ExistingThreadSettingsRouteSession {+ if (!session.engineBotRef || !bot?.engine || bot.archivedAt !== null) return session;++ const selection = botEngineModelSelection(bot.engine);++ const model = models.find(+ (candidate) =>+ candidate.selection.instanceId === selection.instanceId &&+ candidate.selection.model === selection.model,+ );++ if (!model) return session;++ return {+ ...session,+ selectedModel: selection,+ providerGroups: groupByProvider(models).filter(+ (group) => group.providerKey === selection.instanceId,+ ),+ optionDescriptors: resolveProviderOptionDescriptors({+ capabilities: model.capabilities,+ selections: selection.options,+ }),+ onUpdateOptionSelections: (options) =>+ session.onSelectModel({ ...model, selection: { ...selection, options } }),+ };+}diff --git a/apps/mobile/src/features/threads/thread-settings-catalog.tsx b/apps/mobile/src/features/threads/thread-settings-catalog.tsxindex fca4840e5..4e1d2aba7 100644--- a/apps/mobile/src/features/threads/thread-settings-catalog.tsx+++ b/apps/mobile/src/features/threads/thread-settings-catalog.tsx@@ -5,7 +5,7 @@ import { HeaderHeightContext } from "@react-navigation/elements";import { useNavigation } from "@react-navigation/native";import type { NativeStackNavigationProp } from "@react-navigation/native-stack";import { use, useCallback, useMemo, useState, type ReactNode } from "react";-import { Alert, Platform, Pressable, TextInput, View } from "react-native";+import { Platform, Pressable, TextInput, View } from "react-native";import Animated, { FadeIn, FadeOut, LinearTransition } from "react-native-reanimated";import { SymbolView } from "../../components/AppSymbol";import { AppText as Text } from "../../components/AppText";@@ -446,24 +446,12 @@ export function ThreadSettingsModelsScreen() {const navigation = useNavigation<NativeStackNavigationProp<ThreadSettingsPickerStackParams>>();const usesNativeMailSearchToolbar = Platform.OS === "ios" && NATIVE_MAIL_SEARCH_TOOLBAR_SUPPORTED;const hasCustomCatalogFilter = session.providerFilter !== null || session.showLegacy;- const [saving, setSaving] = useState(false);- const hasPendingChanges = session.pendingModel !== null || session.botUsageCapDirty;-- const commitAndClose = useCallback(async () => {- if (!session.botUsageCapValid || saving) return;- setSaving(true);- const saved = await session.commitBotUsageCap();- setSaving(false);-- if (!saved) {- Alert.alert(t("Could not save bot settings"));-- return;- }+ const hasPendingChanges = session.pendingModel !== null;+ const commitAndClose = useCallback(() => {session.commitPendingModel();presentation.onClose();- }, [presentation, saving, session, t]);+ }, [presentation, session]);const filterMenu = useMemo(() => ({@@ -510,9 +498,8 @@ export function ThreadSettingsModelsScreen() {actions={[{accessibilityLabel: hasPendingChanges ? t("Save chat settings") : t("Done"),- disabled: saving || !session.botUsageCapValid,icon: "checkmark",- onPress: () => void commitAndClose(),+ onPress: commitAndClose,},]}onBack={presentation.onClose}@@ -593,9 +580,8 @@ export function ThreadSettingsModelsScreen() {<NativeHeaderToolbar placement="right"><NativeHeaderToolbar.ButtonaccessibilityLabel={hasPendingChanges ? t("Save chat settings") : t("Done")}- disabled={saving || !session.botUsageCapValid}label={hasPendingChanges ? t("Save") : t("Done")}- onPress={() => void commitAndClose()}+ onPress={commitAndClose}/></NativeHeaderToolbar>{Platform.OS === "ios" && !usesNativeMailSearchToolbar ? (diff --git a/apps/mobile/src/features/threads/thread-settings-content.tsx b/apps/mobile/src/features/threads/thread-settings-content.tsxindex f25f1b646..b5491df3f 100644--- a/apps/mobile/src/features/threads/thread-settings-content.tsx+++ b/apps/mobile/src/features/threads/thread-settings-content.tsx@@ -2,7 +2,7 @@ import { useMobileI18n } from "../../lib/i18n";import { getProviderOptionCurrentLabel, getProviderOptionCurrentValue } from "@akeru/shared/model";import * as Haptics from "expo-haptics";import { useState } from "react";-import { Alert, Platform, Pressable, ScrollView, TextInput, View } from "react-native";+import { Alert, Platform, Pressable, ScrollView, View } from "react-native";import Animated, { FadeIn, FadeOut, LinearTransition } from "react-native-reanimated";import { useSafeAreaInsets } from "react-native-safe-area-context";import { SymbolView } from "../../components/AppSymbol";@@ -179,11 +179,7 @@ export function ThreadSettingsOptionsItem(props: {})}<Animated.View layout={THREAD_SETTINGS_OPTIONS_LAYOUT_TRANSITION}><DisclosureRow- isLast={- session.botUsageCapInput === undefined &&- !session.memoryThreadRef &&- !session.routinesRef- }+ isLast={!session.memoryThreadRef && !session.routinesRef}label={t("Runtime")}value={RUNTIME_MODE_CHOICES.find((choice) => choice.mode === session.runtimeMode)?.label@@ -194,7 +190,7 @@ export function ThreadSettingsOptionsItem(props: {{session.memoryThreadRef ? (<Animated.View layout={THREAD_SETTINGS_OPTIONS_LAYOUT_TRANSITION}><DisclosureRow- isLast={session.botUsageCapInput === undefined && !session.routinesRef}+ isLast={!session.routinesRef}label={t("Memory")}value={t("Markdown and observations")}onPress={() => props.onOpenSubmenu({ kind: "memory" })}@@ -204,36 +200,13 @@ export function ThreadSettingsOptionsItem(props: {{session.routinesRef ? (<Animated.View layout={THREAD_SETTINGS_OPTIONS_LAYOUT_TRANSITION}><DisclosureRow- isLast={session.botUsageCapInput === undefined}+ isLastlabel={t("Routines")}value={session.routinesRef.botName}onPress={() => props.onOpenSubmenu({ kind: "routines" })}/></Animated.View>) : null}- {session.botUsageCapInput !== undefined && session.botUsageCapAvailable ? (- <View className="min-h-14 flex-row items-center gap-3 bg-card px-4 py-2">- <Text className="text-sm font-t3-medium text-foreground">{t("Token hard stop")}</Text>- <TextInput- accessibilityLabel={t("Token hard stop")}- className="min-w-24 flex-1 text-right text-base tabular-nums text-foreground"- inputMode="numeric"- keyboardType="number-pad"- onChangeText={session.setBotUsageCapInput}- placeholder={t("No limit")}- placeholderTextColorClassName="accent-placeholder"- returnKeyType="done"- value={session.botUsageCapInput}- />- </View>- ) : session.botUsageCapInput !== undefined ? (- <View className="min-h-14 flex-row items-center justify-between gap-3 bg-card px-4 py-2">- <Text className="text-sm font-t3-medium text-foreground">{t("Token hard stop")}</Text>- <Text className="text-sm text-foreground-muted">- {t("Unavailable for this provider")}- </Text>- </View>- ) : null}</Animated.View>{session.canDelegate ? null : (<Textdiff --git a/apps/mobile/src/features/threads/thread-settings-session.tsx b/apps/mobile/src/features/threads/thread-settings-session.tsxindex a991afa22..db25033b9 100644--- a/apps/mobile/src/features/threads/thread-settings-session.tsx+++ b/apps/mobile/src/features/threads/thread-settings-session.tsx@@ -1,6 +1,5 @@import { driverSupportsDelegation } from "@akeru/shared/delegationProviders";import type {- BotUsageCap,EnvironmentId,ModelSelection,ProviderOptionDescriptor,@@ -13,8 +12,10 @@ import * as Haptics from "expo-haptics";import { createContext, use, useCallback, useMemo, useState, type ReactNode } from "react";import type { ModelOption, ProviderGroup } from "../../lib/modelOptions";import { applyProviderOptionSelection } from "../../lib/providerOptions";-import { pendingModelAfterPress } from "./thread-settings-sheet-state";-import { resolveBotUsageCapForProvider } from "./botStepUsage";+import {+ pendingModelAfterPress,+ stageModelWithAppliedOptions,+} from "./thread-settings-sheet-state";export type ThreadSettingsSubmenuPage =| { readonly kind: "descriptor"; readonly id: string }@@ -30,9 +31,6 @@ export type ThreadSettingsSessionProps = {readonly onUpdateOptionSelections: (selections: ReadonlyArray<ProviderOptionSelection>) => void;readonly runtimeMode: RuntimeMode;readonly onUpdateRuntimeMode: (mode: RuntimeMode) => void;- readonly botUsageCap?: BotUsageCap | null;- readonly botUsageCapProviderDriver?: string;- readonly onUpdateBotUsageCap?: (input: string) => Promise<boolean>;readonly memoryThreadRef?: {readonly environmentId: EnvironmentId;readonly threadId: ThreadId;@@ -49,6 +47,10 @@ export type ThreadSettingsSessionProps = {export type ExistingThreadSettingsRouteSession = ThreadSettingsSessionProps & {readonly ownerId: string;+ readonly engineBotRef?: {+ readonly environmentId: EnvironmentId;+ readonly botId: string;+ };};type ExistingThreadSettingsRouteContextValue = {@@ -97,10 +99,6 @@ export type ThreadSettingsSessionValue = {readonly providerGroups: ReadonlyArray<ProviderGroup>;readonly runtimeMode: RuntimeMode;readonly onUpdateRuntimeMode: (mode: RuntimeMode) => void;- readonly botUsageCapInput: string | undefined;- readonly botUsageCapAvailable: boolean;- readonly botUsageCapDirty: boolean;- readonly botUsageCapValid: boolean;readonly displayedDescriptors: ReadonlyArray<ProviderOptionDescriptor>;readonly providerExpansionOverrides: ReadonlySet<string>;readonly hasLegacyModels: boolean;@@ -110,14 +108,12 @@ export type ThreadSettingsSessionValue = {readonly showLegacy: boolean;readonly applyOptionChange: (id: string, value: string | boolean) => void;readonly commitPendingModel: () => void;- readonly commitBotUsageCap: () => Promise<boolean>;readonly isApplied: (option: ModelOption) => boolean;readonly isDisplayed: (option: ModelOption) => boolean;readonly pressModel: (option: ModelOption) => void;readonly setProviderFilter: (providerKey: string | null) => void;readonly setSearchQuery: (query: string) => void;readonly setShowLegacy: (showLegacy: boolean) => void;- readonly setBotUsageCapInput: (input: string) => void;readonly toggleProvider: (providerKey: string) => void;readonly memoryThreadRef: ThreadSettingsSessionProps["memoryThreadRef"];readonly routinesRef: ThreadSettingsSessionProps["routinesRef"];@@ -142,22 +138,6 @@ export function ThreadSettingsSessionProvider(const [pendingModel, setPendingModel] = useState<ModelOption | null>(null);- const [botUsageCapInput, setBotUsageCapInput] = useState<string | undefined>(() =>- props.botUsageCap === undefined ? undefined : (props.botUsageCap?.limit.toString() ?? ""),- );-- const resolvedBotUsageCap =- botUsageCapInput === undefined- ? undefined- : resolveBotUsageCapForProvider(botUsageCapInput, props.botUsageCapProviderDriver);-- const parsedBotUsageCap = resolvedBotUsageCap?.limit;-- const botUsageCapDirty =- botUsageCapInput !== undefined && parsedBotUsageCap !== (props.botUsageCap?.limit ?? null);-- const botUsageCapValid = botUsageCapInput === undefined || parsedBotUsageCap !== undefined;-const isApplied = useCallback((option: ModelOption) =>option.selection.instanceId === props.selectedModel?.instanceId &&@@ -210,14 +190,6 @@ export function ThreadSettingsSessionProvider(}}, [pendingModel, props.onSelectModel]);- const commitBotUsageCap = useCallback(- () =>- botUsageCapInput === undefined || !botUsageCapDirty- ? Promise.resolve(true)- : (props.onUpdateBotUsageCap?.(botUsageCapInput) ?? Promise.resolve(false)),- [botUsageCapDirty, botUsageCapInput, props.onUpdateBotUsageCap],- );-const applyOptionChange = useCallback((id: string, value: string | boolean) => {const next = applyProviderOptionSelection(displayedDescriptors, { id, value });@@ -257,12 +229,12 @@ export function ThreadSettingsSessionProvider(setPendingModel((current) =>pendingModelAfterPress({current,- pressed: option,+ pressed: stageModelWithAppliedOptions(option, props.selectedModel),pressedIsApplied: isApplied(option),}),);},- [isApplied],+ [isApplied, props.selectedModel],);const value = useMemo<ThreadSettingsSessionValue>(@@ -270,10 +242,6 @@ export function ThreadSettingsSessionProvider(providerGroups: props.providerGroups,runtimeMode: props.runtimeMode,onUpdateRuntimeMode: props.onUpdateRuntimeMode,- botUsageCapInput,- botUsageCapAvailable: resolvedBotUsageCap?.available ?? true,- botUsageCapDirty,- botUsageCapValid,displayedDescriptors,providerExpansionOverrides,hasLegacyModels,@@ -283,14 +251,12 @@ export function ThreadSettingsSessionProvider(showLegacy: showLegacyToggle,applyOptionChange,commitPendingModel,- commitBotUsageCap,isApplied,isDisplayed,pressModel,setProviderFilter,setSearchQuery,setShowLegacy: setShowLegacyToggle,- setBotUsageCapInput,toggleProvider,memoryThreadRef: props.memoryThreadRef,routinesRef: props.routinesRef,@@ -300,11 +266,6 @@ export function ThreadSettingsSessionProvider([applyOptionChange,canDelegate,- botUsageCapDirty,- botUsageCapInput,- botUsageCapValid,- resolvedBotUsageCap?.available,- commitBotUsageCap,commitPendingModel,displayedDescriptors,providerExpansionOverrides,diff --git a/apps/mobile/src/features/threads/thread-settings-sheet-state.test.ts b/apps/mobile/src/features/threads/thread-settings-sheet-state.test.tsindex 5a5b2d852..46867ab38 100644--- a/apps/mobile/src/features/threads/thread-settings-sheet-state.test.ts+++ b/apps/mobile/src/features/threads/thread-settings-sheet-state.test.ts@@ -1,16 +1,26 @@import { describe, expect, it } from "vite-plus/test";-import { ProviderInstanceId, type ProviderOptionSelection } from "@akeru/contracts";+import {+ ProviderInstanceId,+ type ModelCapabilities,+ type ProviderOptionSelection,+} from "@akeru/contracts";import type { ModelOption } from "../../lib/modelOptions";-import { modelMatchesCatalogQuery, pendingModelAfterPress } from "./thread-settings-sheet-state";+import {+ modelMatchesCatalogQuery,+ pendingModelAfterPress,+ stageModelWithAppliedOptions,+} from "./thread-settings-sheet-state";function modelOption(model: string,options: ReadonlyArray<ProviderOptionSelection> = [],+ capabilities: ModelCapabilities | null = null,+ instance = "codex",): ModelOption {return {- key: `codex:${model}`,+ key: `${instance}:${model}`,label: model,subtitle: "Codex",providerKey: "codex",@@ -18,10 +28,10 @@ function modelOption(providerDriver: "codex",isDefault: false,isLegacy: false,- capabilities: null,+ capabilities,disabledReason: null,selection: {- instanceId: ProviderInstanceId.make("codex"),+ instanceId: ProviderInstanceId.make(instance),model,options,},@@ -82,4 +92,43 @@ describe("thread settings sheet state", () => {}),).toBe(pressed);});++ describe("staging a model with the applied options", () => {+ const reasoning = (levels: ReadonlyArray<string>): ModelCapabilities => ({+ optionDescriptors: [+ {+ id: "reasoningEffort",+ label: "Reasoning",+ type: "select",+ currentValue: "default",+ options: [+ { id: "default", label: "Provider default", isDefault: true },+ ...levels.map((id) => ({ id, label: id })),+ ],+ },+ ],+ });++ const applied = modelOption("gpt-a", [{ id: "reasoningEffort", value: "minimal" }]).selection;+ const defaults = [{ id: "reasoningEffort", value: "default" }];++ it("keeps a choice the staged model on the same instance supports", () => {+ const staged = stageModelWithAppliedOptions(+ modelOption("gpt-b", defaults, reasoning(["minimal", "high"])),+ applied,+ );++ expect(staged.selection.options).toEqual([{ id: "reasoningEffort", value: "minimal" }]);+ });++ it("starts on defaults for an unsupported choice, another instance, or unknown options", () => {+ for (const pressed of [+ modelOption("gpt-b", defaults, reasoning(["high"])),+ modelOption("gpt-b", defaults, reasoning(["minimal"]), "codex_work"),+ modelOption("gpt-b", defaults, null),+ ]) {+ expect(stageModelWithAppliedOptions(pressed, applied)).toBe(pressed);+ }+ });+ });});diff --git a/apps/mobile/src/features/threads/thread-settings-sheet-state.ts b/apps/mobile/src/features/threads/thread-settings-sheet-state.tsindex d77db4526..2de05a256 100644--- a/apps/mobile/src/features/threads/thread-settings-sheet-state.ts+++ b/apps/mobile/src/features/threads/thread-settings-sheet-state.ts@@ -1,3 +1,10 @@+import type { ModelSelection } from "@akeru/contracts";+import {+ buildProviderOptionSelectionsFromDescriptors,+ getProviderOptionDescriptors,+ providerOptionsForModelChange,+} from "@akeru/shared/model";+import type { ModelOption } from "../../lib/modelOptions";/** Match the terms a user can actually see or recognize in the model picker. */@@ -33,6 +40,30 @@ export function pendingModelAfterPress(input: {return input.current?.key === input.pressed.key ? input.current : input.pressed;}+/**+ * A model staged on the applied model's instance keeps the applied choices it+ * supports, so Save writes the model and those options together. Another+ * instance, or a model with unknown options, starts on its own defaults.+ */+export function stageModelWithAppliedOptions(+ pressed: ModelOption,+ applied: ModelSelection | null,+): ModelOption {+ const carried = providerOptionsForModelChange({+ previous: applied,+ instanceId: pressed.selection.instanceId,+ nextCaps: pressed.capabilities,+ });++ if (!carried || !pressed.capabilities) return pressed;++ const options = buildProviderOptionSelectionsFromDescriptors(+ getProviderOptionDescriptors({ caps: pressed.capabilities, selections: carried }),+ );++ return options ? { ...pressed, selection: { ...pressed.selection, options } } : pressed;+}+/*** Primary and selected providers start open; all other catalogs start closed.* A user's disclosure tap inverts that default until the picker is dismissed.diff --git a/apps/mobile/src/features/threads/use-thread-composer-settings.ts b/apps/mobile/src/features/threads/use-thread-composer-settings.tsnew file mode 100644index 000000000..d28bd4be3--- /dev/null+++ b/apps/mobile/src/features/threads/use-thread-composer-settings.ts@@ -0,0 +1,165 @@+import { Predicate } from "effect";+import type {+ EnvironmentId,+ ModelSelection,+ OrchestrationBot,+ OrchestrationThreadShell,+ RuntimeMode,+ ServerConfig,+ SubscriptionProviderStatus,+} from "@akeru/contracts";+import { squashAtomCommandFailure } from "@akeru/client-runtime/state/runtime";+import { useCallback, useEffect, useMemo } from "react";++import { useMobileI18n } from "../../lib/i18n";+import { buildModelOptions, groupByProvider } from "../../lib/modelOptions";+import { resolveProviderOptionDescriptors } from "../../lib/providerOptions";+import { botEnvironment } from "../../state/bots";+import { useAtomCommand } from "../../state/use-atom-command";+import type { ExistingThreadSettingsRouteSession } from "./ThreadSettingsSheet";++/**+ * The composer's model and settings sheet: the chat's model row, the session+ * the settings sheet edits, and the action that opens it. Model and option+ * changes go through `onUpdateModelSelection`, which decides whether they land+ * on the chat draft or on a direct bot's engine.+ */+export function useThreadComposerSettings(input: {+ readonly environmentId: EnvironmentId;+ readonly selectedThread: OrchestrationThreadShell;+ readonly serverConfig: ServerConfig | null;+ readonly subscriptionStatuses: ReadonlyArray<SubscriptionProviderStatus> | undefined;+ readonly bot: OrchestrationBot | undefined;+ readonly settingsOwnerId: string;+ readonly settingsRoutePresentation: {+ readonly present: (session: ExistingThreadSettingsRouteSession) => void;+ };+ readonly settingsSheetPresentation: { readonly isActive: boolean; readonly open: () => void };+ readonly onUpdateModelSelection: (modelSelection: ModelSelection) => void;+ readonly onUpdateRuntimeMode: (runtimeMode: RuntimeMode) => void;+}) {+ const {+ environmentId,+ selectedThread,+ serverConfig,+ subscriptionStatuses,+ bot,+ settingsOwnerId,+ onUpdateModelSelection,+ onUpdateRuntimeMode,+ } = input;++ const presentSettingsRoute = input.settingsRoutePresentation.present;+ const settingsSheetActive = input.settingsSheetPresentation.isActive;+ const openSettingsSheet = input.settingsSheetPresentation.open;+ const { t } = useMobileI18n();+ const deleteBot = useAtomCommand(botEnvironment.delete, { reportFailure: false });+ const currentModelSelection = selectedThread.modelSelection;+ const currentRuntimeMode = selectedThread.runtimeMode;++ const modelOptions = useMemo(+ () => buildModelOptions(serverConfig, currentModelSelection, subscriptionStatuses, t),+ [serverConfig, currentModelSelection, subscriptionStatuses, t],+ );++ const providerGroups = useMemo(() => groupByProvider(modelOptions), [modelOptions]);++ // An existing thread is bound to its harness: sessions can't move between+ // provider instances, so the picker only offers the thread's own group.+ const threadProviderGroups = useMemo(+ () => providerGroups.filter((group) => group.providerKey === currentModelSelection.instanceId),+ [providerGroups, currentModelSelection.instanceId],+ );++ const currentModelOption =+ modelOptions.find(+ (option) =>+ option.selection.instanceId === currentModelSelection.instanceId &&+ option.selection.model === currentModelSelection.model,+ ) ?? null;++ const providerOptionDescriptors = useMemo(+ () =>+ resolveProviderOptionDescriptors({+ capabilities: currentModelOption?.capabilities,+ selections: currentModelSelection.options,+ }),+ [currentModelOption?.capabilities, currentModelSelection.options],+ );++ const deleteThreadBot = useCallback(async () => {+ if (!bot) return t("The command failed.");++ const result = await deleteBot({+ environmentId: environmentId,+ input: { botId: bot.id },+ });++ if (!Predicate.isTagged(result, "Failure")) return null;+ const error = squashAtomCommandFailure(result);++ return error instanceof Error ? error.message : t("The command failed.");+ }, [bot, deleteBot, environmentId, t]);++ const settingsRouteSession = useMemo<ExistingThreadSettingsRouteSession>(+ () => ({+ ownerId: settingsOwnerId,+ ...(bot && bot.archivedAt === null && selectedThread.groupId === null+ ? { engineBotRef: { environmentId, botId: bot.id } }+ : {}),+ providerGroups: threadProviderGroups,+ selectedModel: currentModelSelection,+ onSelectModel: (option) => onUpdateModelSelection(option.selection),+ optionDescriptors: providerOptionDescriptors,+ onUpdateOptionSelections: (options) =>+ onUpdateModelSelection({ ...currentModelSelection, options }),+ runtimeMode: currentRuntimeMode,+ onUpdateRuntimeMode: onUpdateRuntimeMode,+ ...(bot+ ? {+ memoryThreadRef: {+ environmentId: environmentId,+ threadId: selectedThread.id,+ },+ routinesRef: {+ environmentId: environmentId,+ botId: bot.id,+ botName: bot.name,+ },+ }+ : {}),+ ...(bot+ ? {+ onDeleteBot: deleteThreadBot,+ }+ : {}),+ }),+ [+ currentModelSelection,+ currentRuntimeMode,+ bot,+ deleteThreadBot,+ environmentId,+ onUpdateModelSelection,+ onUpdateRuntimeMode,+ selectedThread.id,+ selectedThread.groupId,+ providerOptionDescriptors,+ settingsOwnerId,+ threadProviderGroups,+ ],+ );++ const openSettings = useCallback(() => {+ presentSettingsRoute(settingsRouteSession);+ openSettingsSheet();+ }, [presentSettingsRoute, settingsRouteSession, openSettingsSheet]);++ useEffect(() => {+ if (settingsSheetActive) {+ presentSettingsRoute(settingsRouteSession);+ }+ }, [presentSettingsRoute, settingsRouteSession, settingsSheetActive]);++ return { currentModelSelection, currentModelOption, openSettings };+}diff --git a/apps/mobile/src/features/usage/BotUsageDetails.test.tsx b/apps/mobile/src/features/usage/BotUsageDetails.test.tsxindex bef4cc473..8298da2a5 100644--- a/apps/mobile/src/features/usage/BotUsageDetails.test.tsx+++ b/apps/mobile/src/features/usage/BotUsageDetails.test.tsx@@ -19,7 +19,6 @@ const snapshot: AkeruBotUsageSnapshot = {reflector: { tokens: 100, unavailableEntries: 0 },},entries: [],- usageCap: { unit: "tokens", limit: 20_000 },estimatedCost: { status: "available", usd: 1.25 },subscriptionPool: { status: "available", used: 4_000, limit: 10_000, unit: "tokens" },};@@ -68,9 +67,6 @@ describe("mobile bot usage screen", () => {"300","Reflector","100",- "Cap",- "12,345 / 20,000 tokens",- "chat settings","Estimated cost","$1.25","Not subscription spend",@@ -90,7 +86,6 @@ describe("mobile bot usage screen", () => {data: {...snapshot,reservedTokens: 0,- usageCap: null,measurements: {...snapshot.measurements,input: { tokens: 0, unavailableEntries: 1 },@@ -105,7 +100,7 @@ describe("mobile bot usage screen", () => {expect(tree).toContain("Unavailable");expect(tree).toContain("1,250+");- expect(tree).toContain("No cap");+ expect(tree).not.toContain("No cap");expect(tree).toContain("Some provider usage is unavailable.");// Nothing reserved gets no row, matching web.expect(tree).not.toContain("Reserved");diff --git a/apps/mobile/src/features/usage/__snapshots__/BotUsageDetails.test.tsx.snap b/apps/mobile/src/features/usage/__snapshots__/BotUsageDetails.test.tsx.snapindex e68ec48a5..7ebbdc847 100644--- a/apps/mobile/src/features/usage/__snapshots__/BotUsageDetails.test.tsx.snap+++ b/apps/mobile/src/features/usage/__snapshots__/BotUsageDetails.test.tsx.snap@@ -1,3 +1,3 @@// Vitest Snapshot v1, https://vitest.dev/guide/snapshot.html-exports[`mobile bot usage screen > matches the unavailable labelling snapshot 1`] = `"<div class="gap-3" accessibilityLabel="Bot usage"><div class="gap-2"><span class="px-2 text-sm font-t3-medium text-foreground-muted">Scout</span><div class="overflow-hidden rounded-[24px] border-continuous bg-card"><div class="gap-1 p-4"><span class="font-sans text-sm text-foreground-muted">Input</span><span class="font-sans text-lg text-foreground-muted">Unavailable</span></div><div class="gap-1 border-t border-border-subtle p-4"><span class="font-sans text-sm text-foreground-muted">Output</span><span class="font-sans text-lg text-foreground-muted">1,250+</span></div><div class="gap-1 border-t border-border-subtle p-4"><span class="font-sans text-sm text-foreground-muted">Observer</span><span class="font-sans text-lg tabular-nums text-foreground">0</span></div><div class="gap-1 border-t border-border-subtle p-4"><span class="font-sans text-sm text-foreground-muted">Reflector</span><span class="font-sans text-lg text-foreground-muted">Unavailable</span></div><div class="gap-1 border-t border-border-subtle p-4"><span class="font-sans text-sm text-foreground-muted">Cap</span><span class="font-sans text-lg tabular-nums text-foreground">12,345 / 20,000 tokens</span><span class="font-sans text-sm text-foreground-muted">Change the cap in a chat with this bot, under chat settings.</span></div><div class="gap-1 border-t border-border-subtle p-4"><span class="font-sans text-sm text-foreground-muted">Estimated cost</span><span class="font-sans text-lg text-foreground-muted">Unavailable</span><span class="font-sans text-sm text-foreground-muted">Estimated from model rates. Not subscription spend, and not an amount billed.</span></div><div class="gap-1 border-t border-border-subtle p-4"><span class="font-sans text-sm text-foreground-muted">Subscription pool</span><span class="font-sans text-lg text-foreground-muted">Unavailable</span></div><div class="gap-1 border-t border-border-subtle p-4"><span class="font-sans text-sm text-foreground-muted">Reserved</span><span class="font-sans text-lg tabular-nums text-foreground">750</span></div></div></div><div class="rounded-[24px] border-continuous bg-card px-4 py-3"><span class="font-sans text-sm text-foreground-muted">Some provider usage is unavailable.</span></div></div>"`;+exports[`mobile bot usage screen > matches the unavailable labelling snapshot 1`] = `"<div class="gap-3" accessibilityLabel="Bot usage"><div class="gap-2"><span class="px-2 text-sm font-t3-medium text-foreground-muted">Scout</span><div class="overflow-hidden rounded-[24px] border-continuous bg-card"><div class="gap-1 p-4"><span class="font-sans text-sm text-foreground-muted">Input</span><span class="font-sans text-lg text-foreground-muted">Unavailable</span></div><div class="gap-1 border-t border-border-subtle p-4"><span class="font-sans text-sm text-foreground-muted">Output</span><span class="font-sans text-lg text-foreground-muted">1,250+</span></div><div class="gap-1 border-t border-border-subtle p-4"><span class="font-sans text-sm text-foreground-muted">Observer</span><span class="font-sans text-lg tabular-nums text-foreground">0</span></div><div class="gap-1 border-t border-border-subtle p-4"><span class="font-sans text-sm text-foreground-muted">Reflector</span><span class="font-sans text-lg text-foreground-muted">Unavailable</span></div><div class="gap-1 border-t border-border-subtle p-4"><span class="font-sans text-sm text-foreground-muted">Estimated cost</span><span class="font-sans text-lg text-foreground-muted">Unavailable</span><span class="font-sans text-sm text-foreground-muted">Estimated from model rates. Not subscription spend, and not an amount billed.</span></div><div class="gap-1 border-t border-border-subtle p-4"><span class="font-sans text-sm text-foreground-muted">Subscription pool</span><span class="font-sans text-lg text-foreground-muted">Unavailable</span></div><div class="gap-1 border-t border-border-subtle p-4"><span class="font-sans text-sm text-foreground-muted">Reserved</span><span class="font-sans text-lg tabular-nums text-foreground">750</span></div></div></div><div class="rounded-[24px] border-continuous bg-card px-4 py-3"><span class="font-sans text-sm text-foreground-muted">Some provider usage is unavailable.</span></div></div>"`;diff --git a/apps/mobile/src/features/usage/botUsagePresentation.test.ts b/apps/mobile/src/features/usage/botUsagePresentation.test.tsindex 82f8bf12a..ae83cf9de 100644--- a/apps/mobile/src/features/usage/botUsagePresentation.test.ts+++ b/apps/mobile/src/features/usage/botUsagePresentation.test.ts@@ -2,13 +2,11 @@ import { BotId, type AkeruBotUsageSnapshot } from "@akeru/contracts";import { describe, expect, it } from "vite-plus/test";import {- BOT_USAGE_CAP_EDIT_CAPTION,BOT_USAGE_COST_CAPTION,BOT_USAGE_PARTIAL_NOTICE,botUsageView,formatEstimatedCost,formatSubscriptionPool,- formatUsageCap,formatUsageMeasurement,} from "./botUsagePresentation";@@ -23,7 +21,6 @@ const snapshot: AkeruBotUsageSnapshot = {reflector: { tokens: 100, unavailableEntries: 0 },},entries: [],- usageCap: { unit: "tokens", limit: 20_000 },estimatedCost: { status: "available", usd: 1.25 },subscriptionPool: { status: "available", used: 4_000, limit: 10_000, unit: "tokens" },};@@ -67,11 +64,6 @@ describe("bot usage value formatting", () => {formatSubscriptionPool({ status: "unavailable", used: null, limit: null, unit: null }),).toBe("Unavailable");});-- it("shows the cap against consumed tokens, and says so when there is none", () => {- expect(formatUsageCap(snapshot)).toBe("12,345 / 20,000 tokens");- expect(formatUsageCap({ ...snapshot, usageCap: null })).toBe("No cap");- });});describe("botUsageView", () => {@@ -90,7 +82,7 @@ describe("botUsageView", () => {});});- it("builds every measurement, the cap, cost, pool, and reservations", () => {+ it("builds every measurement, cost, pool, and reservations", () => {const view = botUsageView({ data: snapshot, error: null, isPending: false });if (view.kind !== "ready") throw new Error("expected a ready view");@@ -99,7 +91,6 @@ describe("botUsageView", () => {"output","observer","reflector",- "cap","estimated-cost","subscription-pool","reserved",@@ -109,11 +100,10 @@ describe("botUsageView", () => {expect(view.partialNotice).toBeNull();});- it("keeps the estimate separate from subscription spend and states where the cap is edited", () => {+ it("keeps the estimate separate from subscription spend", () => {const view = botUsageView({ data: snapshot, error: null, isPending: false });expect(rowValue(view, "estimated-cost").caption).toBe(BOT_USAGE_COST_CAPTION);expect(BOT_USAGE_COST_CAPTION).toContain("Not subscription spend");- expect(rowValue(view, "cap").caption).toBe(BOT_USAGE_CAP_EDIT_CAPTION);});it("marks partial and unavailable measurements without inventing values", () => {@@ -121,7 +111,6 @@ describe("botUsageView", () => {data: {...snapshot,reservedTokens: 0,- usageCap: null,measurements: {...snapshot.measurements,input: { tokens: 0, unavailableEntries: 1 },@@ -145,7 +134,6 @@ describe("botUsageView", () => {value: "Unavailable",unavailable: true,});- expect(rowValue(view, "cap").value).toBe("No cap");// Nothing reserved gets no row, matching web.if (view.kind !== "ready") throw new Error("expected a ready view");diff --git a/apps/mobile/src/features/usage/botUsagePresentation.ts b/apps/mobile/src/features/usage/botUsagePresentation.tsindex e4a63964f..d070ad6e4 100644--- a/apps/mobile/src/features/usage/botUsagePresentation.ts+++ b/apps/mobile/src/features/usage/botUsagePresentation.ts@@ -78,16 +78,6 @@ export function formatSubscriptionPool(pool: AkeruBotUsageSnapshot["subscriptionreturn `${pool.used.toLocaleString()} / ${pool.limit.toLocaleString()} ${pool.unit}`;}-export function formatUsageCap(snapshot: AkeruBotUsageSnapshot): string {- return snapshot.usageCap === null- ? "No cap"- : `${snapshot.consumedTokens.toLocaleString()} / ${snapshot.usageCap.limit.toLocaleString()} ${snapshot.usageCap.unit}`;-}--/** Where the cap is changed. Mobile has no bot editor; chat settings owns it. */-export const BOT_USAGE_CAP_EDIT_CAPTION =- "Change the cap in a chat with this bot, under chat settings.";-export const BOT_USAGE_COST_CAPTION ="Estimated from model rates. Not subscription spend, and not an amount billed.";@@ -99,13 +89,6 @@ export function buildBotUsageRows(snapshot: AkeruBotUsageSnapshot): ReadonlyArrameasurementRow("output", "Output", snapshot.measurements.output),measurementRow("observer", "Observer", snapshot.measurements.observer),measurementRow("reflector", "Reflector", snapshot.measurements.reflector),- {- key: "cap",- label: "Cap",- value: formatUsageCap(snapshot),- caption: BOT_USAGE_CAP_EDIT_CAPTION,- unavailable: false,- },{key: "estimated-cost",label: "Estimated cost",diff --git a/apps/mobile/src/lib/mobile-fixtures.test-support.ts b/apps/mobile/src/lib/mobile-fixtures.test-support.tsindex 1743bf5d1..81b129886 100644--- a/apps/mobile/src/lib/mobile-fixtures.test-support.ts+++ b/apps/mobile/src/lib/mobile-fixtures.test-support.ts@@ -26,7 +26,6 @@ export function makeMobileBot(overrides: Partial<OrchestrationBot> = {}): Orchesengine: null,sandbox: null,runtimeMode: "full-access",- usageCap: null,imageProvider: null,voiceEnabled: false,channelBindings: [],diff --git a/apps/mobile/src/lib/threadActivity.ts b/apps/mobile/src/lib/threadActivity.tsindex 1a267d58e..a4e7f2f0c 100644--- a/apps/mobile/src/lib/threadActivity.ts+++ b/apps/mobile/src/lib/threadActivity.ts@@ -394,7 +394,6 @@ function botStepMetersEqual(return (left.tokens === right.tokens &&left.costUsd === right.costUsd &&- left.hardStopReached === right.hardStopReached &&(left.engine === right.engine ||(left.engine.provider === right.engine.provider &&left.engine.model === right.engine.model &&diff --git a/apps/mobile/src/lib/threadWorkLog.ts b/apps/mobile/src/lib/threadWorkLog.tsindex 7e08d4cb7..48012001c 100644--- a/apps/mobile/src/lib/threadWorkLog.ts+++ b/apps/mobile/src/lib/threadWorkLog.ts@@ -104,8 +104,6 @@ export function deriveWorkLogEntries(if (activity.kind === "bot.step-usage.updated") continue;- if (activity.kind === "bot.usage-cap.hit") continue;-// Silent-run state drives the status line; it is not work the bot did.if (isSilentRunActivity(activity)) continue;diff --git a/apps/mobile/src/state/thread-bot-engine.test.ts b/apps/mobile/src/state/thread-bot-engine.test.tsnew file mode 100644index 000000000..553282e5d--- /dev/null+++ b/apps/mobile/src/state/thread-bot-engine.test.ts@@ -0,0 +1,146 @@+import { Schema } from "effect";+import {+ BotId,+ GroupId,+ ProviderInstanceId,+ ServerProvider,+ type ModelCapabilities,+ type OrchestrationBot,+} from "@akeru/contracts";+import { describe, expect, it } from "vite-plus/test";++import { makeMobileBot, makeMobileServerConfig } from "../lib/mobile-fixtures.test-support";+import {+ effectiveThreadModelSelection,+ threadBotEngineUpdate,+ threadEngineBot,+} from "./thread-bot-engine";++const codex = ProviderInstanceId.make("codex");++function bot(overrides: Partial<OrchestrationBot> = {}): OrchestrationBot {+ return makeMobileBot({+ id: BotId.make("bot-1"),+ name: "Mira",+ archivedAt: null,+ engine: {+ provider: "codex",+ model: "gpt-a",+ options: [{ id: "reasoningEffort", value: "minimal" }],+ },+ ...overrides,+ });+}++const reasoning: ModelCapabilities = {+ optionDescriptors: [+ {+ id: "reasoningEffort",+ label: "Reasoning",+ type: "select",+ currentValue: "default",+ options: [+ { id: "default", label: "Provider default", isDefault: true },+ { id: "minimal", label: "Minimal" },+ { id: "high", label: "High" },+ ],+ },+ ],+};++const config = makeMobileServerConfig({+ providers: [+ Schema.decodeUnknownSync(ServerProvider)({+ instanceId: "codex",+ driver: "codex",+ enabled: true,+ installed: true,+ version: null,+ status: "ready",+ auth: { status: "authenticated" },+ checkedAt: "2026-01-01T00:00:00Z",+ slashCommands: [],+ skills: [],+ models: [+ { slug: "gpt-a", name: "A", isCustom: false, capabilities: reasoning },+ { slug: "gpt-custom", name: "Custom", isCustom: true, capabilities: null },+ ],+ }),+ ],+});++describe("thread bot engine", () => {+ it("targets the live bot of a direct chat", () => {+ const mira = bot();++ expect(threadEngineBot({ botId: mira.id, groupId: null }, [mira])).toBe(mira);+ });++ it("never targets a bot from a group, plain, or archived chat", () => {+ const mira = bot();++ expect(+ threadEngineBot({ botId: mira.id, groupId: GroupId.make("group-1") }, [mira]),+ ).toBeNull();+ expect(threadEngineBot({ botId: null, groupId: null }, [mira])).toBeNull();+ expect(+ threadEngineBot({ botId: mira.id }, [bot({ archivedAt: "2026-01-02T00:00:00Z" })]),+ ).toBeNull();+ });++ it("shows and sends the bot's engine over a stale draft", () => {+ const draftSelection = { instanceId: codex, model: "gpt-old" };+ const threadSelection = { instanceId: codex, model: "gpt-thread" };++ expect(effectiveThreadModelSelection({ bot: bot(), draftSelection, threadSelection })).toEqual({+ instanceId: codex,+ model: "gpt-a",+ options: [{ id: "reasoningEffort", value: "minimal" }],+ });+ expect(+ effectiveThreadModelSelection({+ bot: bot({ engine: null }),+ draftSelection,+ threadSelection,+ }),+ ).toBe(draftSelection);+ expect(+ effectiveThreadModelSelection({ bot: null, draftSelection: null, threadSelection }),+ ).toBe(threadSelection);+ });++ it("saves a model and its options to the bot in one engine", () => {+ expect(+ threadBotEngineUpdate(config, {+ instanceId: codex,+ model: "gpt-a",+ options: [{ id: "reasoningEffort", value: "high" }],+ }),+ ).toEqual({+ provider: "codex",+ model: "gpt-a",+ options: [{ id: "reasoningEffort", value: "high" }],+ });+ });++ it("drops a provider default and keeps options on a model with unknown capabilities", () => {+ expect(+ threadBotEngineUpdate(config, {+ instanceId: codex,+ model: "gpt-a",+ options: [{ id: "reasoningEffort", value: "default" }],+ }),+ ).toEqual({ provider: "codex", model: "gpt-a" });+ expect(+ threadBotEngineUpdate(config, {+ instanceId: codex,+ model: "gpt-custom",+ options: [{ id: "reasoningEffort", value: "high" }],+ }),+ ).toEqual({+ provider: "codex",+ model: "gpt-custom",+ options: [{ id: "reasoningEffort", value: "high" }],+ });+ });+});diff --git a/apps/mobile/src/state/thread-bot-engine.ts b/apps/mobile/src/state/thread-bot-engine.tsnew file mode 100644index 000000000..ebece3909--- /dev/null+++ b/apps/mobile/src/state/thread-bot-engine.ts@@ -0,0 +1,60 @@+import {+ type BotEngine,+ type ModelCapabilities,+ type ModelSelection,+ type OrchestrationBot,+ type ServerConfig,+} from "@akeru/contracts";+import { botEngineFromModelSelection, botEngineModelSelection } from "@akeru/shared/model";++export type ThreadBotRef = {+ readonly botId?: string | null | undefined;+ readonly groupId?: string | null | undefined;+};++/**+ * The bot whose engine a chat runs on: the live bot of a direct chat. Group+ * chats pick a responder per turn and plain chats have no bot, so both keep+ * the thread's own selection and must never write a bot's engine.+ */+export function threadEngineBot(+ thread: ThreadBotRef | null | undefined,+ bots: ReadonlyArray<OrchestrationBot>,+): OrchestrationBot | null {+ if (!thread?.botId || (thread.groupId ?? null) !== null) return null;++ return bots.find((bot) => bot.id === thread.botId && bot.archivedAt === null) ?? null;+}++/**+ * The selection a chat shows and sends with. A direct bot's saved engine wins+ * over a stale draft, as the server does when the turn starts; a bot without+ * an engine runs whatever the chat has selected.+ */+export function effectiveThreadModelSelection(input: {+ readonly bot: OrchestrationBot | null;+ readonly draftSelection: ModelSelection | null | undefined;+ readonly threadSelection: ModelSelection;+}): ModelSelection {+ return input.bot?.engine+ ? botEngineModelSelection(input.bot.engine)+ : (input.draftSelection ?? input.threadSelection);+}++/** What the server says a model's options are; null or undefined when it cannot say. */+export function serverModelCapabilities(+ config: ServerConfig | null | undefined,+ selection: ModelSelection,+): ModelCapabilities | null | undefined {+ return config?.providers+ .find((provider) => provider.instanceId === selection.instanceId)+ ?.models.find((model) => model.slug === selection.model)?.capabilities;+}++/** The engine a direct bot saves for a picked selection, keeping only chosen options. */+export function threadBotEngineUpdate(+ config: ServerConfig | null | undefined,+ selection: ModelSelection,+): BotEngine {+ return botEngineFromModelSelection(selection, serverModelCapabilities(config, selection));+}diff --git a/apps/mobile/src/state/use-thread-composer-state.ts b/apps/mobile/src/state/use-thread-composer-state.tsindex 2946ffabe..05c313e09 100644--- a/apps/mobile/src/state/use-thread-composer-state.ts+++ b/apps/mobile/src/state/use-thread-composer-state.ts@@ -58,6 +58,9 @@ import { enqueueThreadOutboxMessage } from "./thread-outbox";import { useThreadOutboxMessages } from "./use-thread-outbox";import { threadEnvironment } from "./threads";import { useAtomCommand } from "./use-atom-command";+import { botEnvironment } from "./bots";+import { effectiveThreadModelSelection, threadBotEngineUpdate } from "./thread-bot-engine";+import { useThreadEngineBot } from "./use-thread-engine-bot";const decodeFeedDelegations = Schema.decodeUnknownSync(Schema.Struct({@@ -116,6 +119,8 @@ export function useThreadComposerState() {reportFailure: false,});+ const updateBot = useAtomCommand(botEnvironment.update, { reportFailure: false });+useEffect(() => {ensureComposerDraftsLoaded();}, []);@@ -216,7 +221,17 @@ export function useThreadComposerState() {const selectedDraft = useComposerDraftSettings(selectedThreadKey);const selectedThreadQueueCount = selectedThreadQueuedMessages.length;const selectedThread = selectedThreadDetail ?? selectedThreadShell;- const modelSelection = selectedDraft.modelSelection ?? selectedThread?.modelSelection ?? null;+ // A direct bot chat shows and sends with the bot's engine, never a stale draft.+ const engineBot = useThreadEngineBot(selectedThreadShell?.environmentId, selectedThread);++ const modelSelection = selectedThread+ ? effectiveThreadModelSelection({+ bot: engineBot,+ draftSelection: selectedDraft.modelSelection,+ threadSelection: selectedThread.modelSelection,+ })+ : null;+const runtimeMode = selectedDraft.runtimeMode ?? selectedThread?.runtimeMode ?? null;const selectedThreadSessionActivity = useMemo(() => {@@ -351,7 +366,11 @@ export function useThreadComposerState() {commandId: CommandId.make(metadata.commandId),text,attachments,- modelSelection: draft.modelSelection ?? thread.modelSelection,+ modelSelection: effectiveThreadModelSelection({+ bot: engineBot,+ draftSelection: draft.modelSelection,+ threadSelection: thread.modelSelection,+ }),runtimeMode: draft.runtimeMode ?? thread.runtimeMode,interactionMode: DEFAULT_PROVIDER_INTERACTION_MODE,createdAt: metadata.createdAt,@@ -372,6 +391,7 @@ export function useThreadComposerState() {return messageId;}, [+ engineBot,selectedEnvironmentRuntime?.serverConfig?.providers,selectedThreadDetail,selectedThreadShell,@@ -476,15 +496,40 @@ export function useThreadComposerState() {[selectedThreadShell],);+ // A direct bot chat saves the pick to the bot, model and options together;+ // every other chat keeps it on its own draft.const onUpdateModelSelection = useCallback((value: ModelSelection) => {- if (!selectedThreadKey) {+ if (!selectedThreadKey || !selectedThreadShell) {+ return;+ }++ if (!engineBot) {+ updateComposerDraftSettings(selectedThreadKey, { modelSelection: value });+return;}- updateComposerDraftSettings(selectedThreadKey, { modelSelection: value });+ void updateBot({+ environmentId: selectedThreadShell.environmentId,+ input: {+ botId: engineBot.id,+ engine: threadBotEngineUpdate(selectedEnvironmentRuntime?.serverConfig, value),+ },+ }).then((result) => {+ if (Predicate.isTagged(result, "Failure") && !isAtomCommandInterrupted(result)) {+ Alert.alert(t("Could not change the model"));+ }+ });},- [selectedThreadKey],+ [+ engineBot,+ selectedEnvironmentRuntime?.serverConfig,+ selectedThreadKey,+ selectedThreadShell,+ t,+ updateBot,+ ],);const onUpdateRuntimeMode = useCallback(diff --git a/apps/mobile/src/state/use-thread-engine-bot.test.ts b/apps/mobile/src/state/use-thread-engine-bot.test.tsnew file mode 100644index 000000000..cf981445c--- /dev/null+++ b/apps/mobile/src/state/use-thread-engine-bot.test.ts@@ -0,0 +1,34 @@+import { BotId, EnvironmentId, type OrchestrationBot } from "@akeru/contracts";+import { beforeEach, describe, expect, it, vi } from "vite-plus/test";+import { makeMobileBot } from "../lib/mobile-fixtures.test-support";+import { environmentBotsAtom } from "./bots";+import { useThreadEngineBot } from "./use-thread-engine-bot";++const state = vi.hoisted(() => ({ bots: [] as ReadonlyArray<OrchestrationBot> }));++vi.mock("@effect/atom-react", () => ({ useAtomValue: () => state.bots }));++vi.mock("react", () => ({ useMemo: <T>(factory: () => T) => factory() }));++vi.mock("./bots", () => ({ environmentBotsAtom: vi.fn(() => ({})) }));++beforeEach(() => {+ state.bots = [];+ vi.clearAllMocks();+});++describe("composer engine bot lookup", () => {+ it("loads and renders without a selected environment", () => {+ expect(useThreadEngineBot(undefined, undefined)).toBeNull();+ expect(environmentBotsAtom).not.toHaveBeenCalled();+ });++ it("looks up the direct chat bot in the selected environment", () => {+ const environmentId = EnvironmentId.make("environment-1");+ const bot = makeMobileBot({ id: BotId.make("scout"), archivedAt: null });+ state.bots = [bot];++ expect(useThreadEngineBot(environmentId, { botId: bot.id, groupId: null })).toBe(bot);+ expect(environmentBotsAtom).toHaveBeenCalledWith(environmentId);+ });+});diff --git a/apps/mobile/src/state/use-thread-engine-bot.ts b/apps/mobile/src/state/use-thread-engine-bot.tsnew file mode 100644index 000000000..8b990e2cd--- /dev/null+++ b/apps/mobile/src/state/use-thread-engine-bot.ts@@ -0,0 +1,21 @@+import { useAtomValue } from "@effect/atom-react";+import { type EnvironmentId, type OrchestrationBot } from "@akeru/contracts";+import { Atom } from "effect/unstable/reactivity";+import { useMemo } from "react";++import { environmentBotsAtom } from "./bots";+import { threadEngineBot, type ThreadBotRef } from "./thread-bot-engine";++const NO_BOTS = Atom.make<ReadonlyArray<OrchestrationBot>>([]);++/** The live bot whose engine the selected chat runs on, following environment snapshots. */+export function useThreadEngineBot(+ environmentId: EnvironmentId | null | undefined,+ thread: ThreadBotRef | null | undefined,+): OrchestrationBot | null {+ const bots = useAtomValue(environmentId ? environmentBotsAtom(environmentId) : NO_BOTS);+ const botId = thread?.botId ?? null;+ const groupId = thread?.groupId ?? null;++ return useMemo(() => threadEngineBot({ botId, groupId }, bots), [botId, bots, groupId]);+}diff --git a/apps/server/src/channels/ChannelInboundRestart.test.ts b/apps/server/src/channels/ChannelInboundRestart.test.tsindex ba88bd7ce..ad3d5e693 100644--- a/apps/server/src/channels/ChannelInboundRestart.test.ts+++ b/apps/server/src/channels/ChannelInboundRestart.test.ts@@ -115,7 +115,6 @@ const seedProjectsAndBot = Effect.fn("seedProjectsAndBot")(function* (root: striengine: null,sandbox: "local",runtimeMode: "full-access",- usageCap: null,groupId: null,createdAt: NOW,});diff --git a/apps/server/src/channels/testUtils/channelRuntime.ts b/apps/server/src/channels/testUtils/channelRuntime.tsindex d4eb65ee8..b3f6a0e60 100644--- a/apps/server/src/channels/testUtils/channelRuntime.ts+++ b/apps/server/src/channels/testUtils/channelRuntime.ts@@ -303,7 +303,6 @@ function makeBot(engine: null,sandbox: "local",runtimeMode: "full-access",- usageCap: null,imageProvider: null,voiceEnabled: false,channelBindings: input.channelBindings ?? [],diff --git a/apps/server/src/channels/testUtils/channelTransports.ts b/apps/server/src/channels/testUtils/channelTransports.tsindex 76ad4bc37..59d474aa6 100644--- a/apps/server/src/channels/testUtils/channelTransports.ts+++ b/apps/server/src/channels/testUtils/channelTransports.ts@@ -190,7 +190,6 @@ const makeBot = (): OrchestrationBot => ({engine: null,sandbox: "local",runtimeMode: "full-access",- usageCap: null,imageProvider: null,voiceEnabled: false,channelBindings: [],diff --git a/apps/server/src/image-generation/testUtils/imageGenerationRuntime.ts b/apps/server/src/image-generation/testUtils/imageGenerationRuntime.tsindex 4dba98799..411e42c65 100644--- a/apps/server/src/image-generation/testUtils/imageGenerationRuntime.ts+++ b/apps/server/src/image-generation/testUtils/imageGenerationRuntime.ts@@ -184,7 +184,6 @@ const createBot = (engine: { provider, model: provider === "codex" ? "gpt-5.6" : "claude-opus-5.5" },sandbox: "local",runtimeMode: "full-access",- usageCap: null,groupId: null,createdAt,});diff --git a/apps/server/src/orchestration/Layers/BotPersistence.ownership.test.ts b/apps/server/src/orchestration/Layers/BotPersistence.ownership.test.tsindex 3768ee4d7..5b062da30 100644--- a/apps/server/src/orchestration/Layers/BotPersistence.ownership.test.ts+++ b/apps/server/src/orchestration/Layers/BotPersistence.ownership.test.ts@@ -44,7 +44,6 @@ it.layer(TestLayer)("bot persistence", (it) => {engine: null,sandbox: "local",runtimeMode: "full-access",- usageCap: null,groupId: input.groupId,createdAt,});@@ -171,7 +170,6 @@ it.layer(TestLayer)("bot persistence", (it) => {engine: null,sandbox: "local",runtimeMode: "full-access",- usageCap: null,groupId: null,createdAt,});@@ -277,7 +275,6 @@ it.layer(TestLayer)("bot persistence", (it) => {engine: null,sandbox: "local",runtimeMode: "full-access",- usageCap: null,groupId: null,createdAt,});@@ -291,7 +288,6 @@ it.layer(TestLayer)("bot persistence", (it) => {engine: null,sandbox: "local",runtimeMode: "full-access",- usageCap: null,groupId: null,createdAt,});diff --git a/apps/server/src/orchestration/Layers/BotPersistence.test.ts b/apps/server/src/orchestration/Layers/BotPersistence.test.tsindex f1cfb084b..3bc1a64e2 100644--- a/apps/server/src/orchestration/Layers/BotPersistence.test.ts+++ b/apps/server/src/orchestration/Layers/BotPersistence.test.ts@@ -40,7 +40,6 @@ it.layer(TestLayer)("bot persistence", (it) => {engine: { provider: "codex", model: "gpt-5.6" },sandbox: "local",runtimeMode: "full-access",- usageCap: null,groupId: null,createdAt,});@@ -95,7 +94,6 @@ it.layer(TestLayer)("bot persistence", (it) => {},sandbox: null,runtimeMode: "approval-required",- usageCap: { unit: "tokens", limit: 50_000 },imageProvider: "chatgpt",personalityTone: 50,voiceEnabled: true,@@ -129,7 +127,6 @@ it.layer(TestLayer)("bot persistence", (it) => {engine: null,sandbox: "local",runtimeMode: "full-access",- usageCap: null,groupId: null,createdAt,});@@ -172,7 +169,6 @@ it.layer(TestLayer)("bot persistence", (it) => {},sandbox: null,runtimeMode: "approval-required",- usageCap: { unit: "tokens", limit: 50_000 },imageProvider: "chatgpt",personalityTone: 50,voiceEnabled: true,@@ -249,7 +245,6 @@ it.layer(TestLayer)("bot persistence", (it) => {avatar: { kind: "dither", seed: "painter" },engine: { provider: "claudeAgent", model: "claude-opus-5.5" },sandbox: "local",- usageCap: null,groupId: null,createdAt,});@@ -292,7 +287,6 @@ it.layer(TestLayer)("bot persistence", (it) => {avatar: { kind: "dither", seed: "akeru" },engine: null,sandbox,- usageCap: null,groupId: null,createdAt: "2026-01-01T00:00:00.000Z",});@@ -331,7 +325,6 @@ it.layer(TestLayer)("bot persistence", (it) => {avatar: { kind: "dither", seed: "legacy" },engine: null,sandbox: "local",- usageCap: null,groupId: null,createdAt: "2026-01-01T00:00:00.000Z",});diff --git a/apps/server/src/orchestration/Layers/DelegationPersistence.test.ts b/apps/server/src/orchestration/Layers/DelegationPersistence.test.tsindex 653585d0a..7fc87f1be 100644--- a/apps/server/src/orchestration/Layers/DelegationPersistence.test.ts+++ b/apps/server/src/orchestration/Layers/DelegationPersistence.test.ts@@ -110,7 +110,6 @@ it.effect("rebuilds the full delegation record after a restart", () =>engine: null,sandbox: "local",runtimeMode: "approval-required",- usageCap: null,groupId: null,createdAt: NOW,});@@ -195,7 +194,6 @@ const seedParentAndChild = Effect.gen(function* () {engine: null,sandbox: "local",runtimeMode: "approval-required",- usageCap: null,groupId: null,createdAt: NOW,});diff --git a/apps/server/src/orchestration/Layers/ProjectionSnapshotRows.ts b/apps/server/src/orchestration/Layers/ProjectionSnapshotRows.tsindex 87f5d4e7a..ad2530b0c 100644--- a/apps/server/src/orchestration/Layers/ProjectionSnapshotRows.ts+++ b/apps/server/src/orchestration/Layers/ProjectionSnapshotRows.ts@@ -4,7 +4,6 @@ import {BotAvatar,BotEngine,BotId,- BotUsageCap,GroupId,ChannelBinding,ChannelDeliveryState,@@ -86,7 +85,6 @@ export const ProjectionBotDbRowSchema = ProjectionBot.mapFields(Struct.assign({avatar: Schema.fromJsonString(BotAvatar),engine: Schema.NullOr(Schema.fromJsonString(BotEngine)),- usageCap: Schema.NullOr(Schema.fromJsonString(BotUsageCap)),personalityTone: Schema.Number,disabledMcpServerIds: Schema.fromJsonString(Schema.Array(McpServerId)),channelBindings: Schema.fromJsonString(Schema.Array(ChannelBinding)),@@ -472,7 +470,6 @@ export function mapBotRow(engine: row.engine,sandbox: row.sandbox,runtimeMode: row.runtimeMode,- usageCap: row.usageCap,imageProvider: row.imageProvider,personalityTone: row.personalityTone,voiceEnabled: row.voiceEnabled === 1,diff --git a/apps/server/src/orchestration/Layers/ProviderCommandReactor.delegation.test.ts b/apps/server/src/orchestration/Layers/ProviderCommandReactor.delegation.test.tsindex 4cd644d97..21eeb883a 100644--- a/apps/server/src/orchestration/Layers/ProviderCommandReactor.delegation.test.ts+++ b/apps/server/src/orchestration/Layers/ProviderCommandReactor.delegation.test.ts@@ -39,7 +39,6 @@ describe("ProviderCommandReactor", () => {engine: null,sandbox: "local",runtimeMode: "approval-required",- usageCap: null,groupId: null,createdAt: now,});@@ -182,7 +181,6 @@ describe("ProviderCommandReactor", () => {engine: null,sandbox: "local",runtimeMode: "approval-required",- usageCap: null,groupId: null,createdAt: now,});@@ -317,7 +315,6 @@ describe("ProviderCommandReactor", () => {engine: null,sandbox: "local",runtimeMode: "approval-required",- usageCap: null,groupId: null,createdAt: now,});diff --git a/apps/server/src/orchestration/Layers/ProviderCommandReactor.delegationRelease.test.ts b/apps/server/src/orchestration/Layers/ProviderCommandReactor.delegationRelease.test.tsindex a4a96e2ae..88d78e34b 100644--- a/apps/server/src/orchestration/Layers/ProviderCommandReactor.delegationRelease.test.ts+++ b/apps/server/src/orchestration/Layers/ProviderCommandReactor.delegationRelease.test.ts@@ -45,7 +45,6 @@ describe("ProviderCommandReactor", () => {engine: null,sandbox: "local",runtimeMode: "approval-required",- usageCap: null,groupId: null,createdAt: now,});@@ -184,7 +183,6 @@ describe("ProviderCommandReactor", () => {engine: null,sandbox: "local",runtimeMode: "approval-required",- usageCap: null,groupId: null,createdAt: now,});@@ -328,7 +326,6 @@ describe("ProviderCommandReactor", () => {engine: null,sandbox: "local",runtimeMode: "approval-required",- usageCap: null,groupId: null,createdAt: now,});@@ -481,7 +478,6 @@ describe("ProviderCommandReactor", () => {engine: null,sandbox: "local",runtimeMode: "approval-required",- usageCap: null,groupId: null,createdAt: now,});diff --git a/apps/server/src/orchestration/Layers/ProviderCommandReactor.turnStart.test.ts b/apps/server/src/orchestration/Layers/ProviderCommandReactor.turnStart.test.tsindex 21741f85e..0aa23701a 100644--- a/apps/server/src/orchestration/Layers/ProviderCommandReactor.turnStart.test.ts+++ b/apps/server/src/orchestration/Layers/ProviderCommandReactor.turnStart.test.ts@@ -5,6 +5,7 @@ import * as Deferred from "effect/Deferred";import { it as effectIt } from "@effect/vitest";import { afterEach, describe, expect, it } from "vite-plus/test";import { ProviderAdapterRequestError } from "../../provider/Errors.ts";+import { AKERU_TURN_USAGE_RESERVATION_TOKENS } from "../../usage/BotUsageLedger.ts";import {asMessageId,createProviderCommandHarness,@@ -120,12 +121,11 @@ describe("ProviderCommandReactor", () => {});});- effectIt.effect("reserves a configured bot's usage cap and binds the provider turn", () =>+ effectIt.effect("reserves a configured bot's turn usage and binds the provider turn", () =>Effect.gen(function* () {const harness = yield* Effect.promise(() =>createHarness({botEngine: { provider: "codex", model: "gpt-5-codex" },- botUsageCap: { unit: "tokens", limit: 1_000 },}),);@@ -151,59 +151,14 @@ describe("ProviderCommandReactor", () => {),);const usage = yield* Effect.promise(() => harness.summarizeBotUsage());- expect(usage.reservedTokens).toBe(1_000);+ expect(usage.reservedTokens).toBe(AKERU_TURN_USAGE_RESERVATION_TOKENS);expect(usage.entries).toContainEqual(expect.objectContaining({botId: "bot-1",threadId: "thread-1",turnId: "turn-1",state: "reserved",- reservedTokens: 1_000,- }),- );- }),- );-- effectIt.effect("rejects a new turn after a configured bot reserves its full usage cap", () =>- Effect.gen(function* () {- const harness = yield* Effect.promise(() =>- createHarness({- botEngine: { provider: "codex", model: "gpt-5-codex" },- botUsageCap: { unit: "tokens", limit: 1_000 },- }),- );-- const dispatchTurn = (suffix: string) =>- harness.engine.dispatch({- type: "thread.turn.start",- commandId: CommandId.make(`cmd-turn-start-cap-${suffix}`),- threadId: ThreadId.make("thread-1"),- message: {- messageId: asMessageId(`user-message-cap-${suffix}`),- role: "user" as const,- text: suffix,- attachments: [],- },- interactionMode: DEFAULT_PROVIDER_INTERACTION_MODE,- runtimeMode: "approval-required" as const,- createdAt: `2026-01-01T00:00:0${suffix === "first" ? "0" : "1"}.000Z`,- });-- yield* dispatchTurn("first");- yield* Effect.promise(() => harness.waitFor(() => harness.sendTurn.mock.calls.length === 1));- yield* dispatchTurn("second");- yield* Effect.promise(() => harness.drain());-- expect(harness.sendTurn).toHaveBeenCalledTimes(1);- const readModel = yield* Effect.promise(() => harness.readModel());- const thread = readModel.threads.find((entry) => entry.id === ThreadId.make("thread-1"));- expect(thread?.activities).toContainEqual(- expect.objectContaining({- kind: "provider.turn.start.failed",- payload: expect.objectContaining({- detail: "Bot bot-1 reached its 1000-token usage cap.",- requestId: "user-message-cap-second",- }),+ reservedTokens: AKERU_TURN_USAGE_RESERVATION_TOKENS,}),);}),@@ -214,7 +169,6 @@ describe("ProviderCommandReactor", () => {const harness = yield* Effect.promise(() =>createHarness({botEngine: { provider: "codex", model: "gpt-5-codex" },- botUsageCap: { unit: "tokens", limit: 1_000 },}),);@@ -248,7 +202,6 @@ describe("ProviderCommandReactor", () => {const harness = yield* Effect.promise(() =>createHarness({botEngine: { provider: "codex", model: "gpt-5-codex" },- botUsageCap: { unit: "tokens", limit: 1_000 },bindTurnFailure: true,}),);@@ -271,7 +224,7 @@ describe("ProviderCommandReactor", () => {const usage = yield* Effect.promise(() => harness.summarizeBotUsage());expect(harness.sendTurn).toHaveBeenCalledTimes(1);- expect(usage.consumedTokens).toBe(1_000);+ expect(usage.consumedTokens).toBe(AKERU_TURN_USAGE_RESERVATION_TOKENS);expect(usage.reservedTokens).toBe(0);expect(usage.entries[0]).toMatchObject({state: "unavailable",diff --git a/apps/server/src/orchestration/Layers/ProviderRuntimeIngestion.channels.test.ts b/apps/server/src/orchestration/Layers/ProviderRuntimeIngestion.channels.test.tsindex 1bbafe1de..4841af610 100644--- a/apps/server/src/orchestration/Layers/ProviderRuntimeIngestion.channels.test.ts+++ b/apps/server/src/orchestration/Layers/ProviderRuntimeIngestion.channels.test.ts@@ -244,7 +244,6 @@ describe("ProviderRuntimeIngestion", () => {engine: { provider: ProviderInstanceId.make("codex"), model: "gpt-5-codex" },sandbox: null,runtimeMode: "approval-required",- usageCap: null,groupId: null,createdAt,});diff --git a/apps/server/src/orchestration/Layers/projection/Bots.ts b/apps/server/src/orchestration/Layers/projection/Bots.tsindex 5ab6f3ac1..3dfcb2f11 100644--- a/apps/server/src/orchestration/Layers/projection/Bots.ts+++ b/apps/server/src/orchestration/Layers/projection/Bots.ts@@ -20,7 +20,6 @@ export function createBots({engine: event.payload.engine,sandbox: event.payload.sandbox,runtimeMode: event.payload.runtimeMode,- usageCap: event.payload.usageCap,imageProvider: event.payload.imageProvider,personalityTone: event.payload.personalityTone,voiceEnabled: event.payload.voiceEnabled,@@ -53,7 +52,6 @@ export function createBots({...(event.payload.runtimeMode !== undefined? { runtimeMode: event.payload.runtimeMode }: {}),- ...(event.payload.usageCap !== undefined ? { usageCap: event.payload.usageCap } : {}),...(event.payload.imageProvider !== undefined? { imageProvider: event.payload.imageProvider }: {}),diff --git a/apps/server/src/orchestration/Layers/provider-command/Failures.ts b/apps/server/src/orchestration/Layers/provider-command/Failures.tsindex 3736dbb9b..9dc2bab02 100644--- a/apps/server/src/orchestration/Layers/provider-command/Failures.ts+++ b/apps/server/src/orchestration/Layers/provider-command/Failures.ts@@ -13,7 +13,6 @@ import { readableErrorDetail } from "../../../provider/Errors.ts";import { providerUnavailabilityFromDetail } from "../../../provider/providerSnapshot.ts";import {resolveControllerBotId,- isBotUsageCapExceeded,isComposioOperationError,isProviderAdapterRequestError,isUnsupportedEngineError,@@ -201,9 +200,6 @@ export function createFailures({const formatFailureDetail = (cause: Cause.Cause<unknown>): string => {const failReason = cause.reasons.find(Cause.isFailReason);- const capError = isBotUsageCapExceeded(failReason?.error) ? failReason.error : undefined;-- if (capError) return capError.message;const composioError = isComposioOperationError(failReason?.error)? failReason.error@@ -218,10 +214,6 @@ export function createFailures({const detail = formatFailureDetail(cause);const failReason = cause.reasons.find(Cause.isFailReason);- if (isBotUsageCapExceeded(failReason?.error)) {- return { detail, unavailability: "usage-cap" };- }-// An engine refusal names the instance that failed, so the chat can// repair that one even after the bot moves to another model.if (isUnsupportedEngineError(failReason?.error)) {diff --git a/apps/server/src/orchestration/Layers/provider-command/Fields.ts b/apps/server/src/orchestration/Layers/provider-command/Fields.tsindex 6865faf46..e332e34df 100644--- a/apps/server/src/orchestration/Layers/provider-command/Fields.ts+++ b/apps/server/src/orchestration/Layers/provider-command/Fields.ts@@ -15,14 +15,11 @@ import {ProviderAdapterRequestError,} from "../../../provider/Errors.ts";import type { AgentControllerError } from "../../../provider/Errors.ts";-import { BotUsageCapExceeded } from "../../../usage/BotUsageLedger.ts";export const isProviderAdapterRequestError = Schema.is(ProviderAdapterRequestError);export const isUnsupportedEngineError = Schema.is(AgentControllerUnsupportedEngineError);-export const isBotUsageCapExceeded = Schema.is(BotUsageCapExceeded);-export const isComposioOperationError = Schema.is(ComposioOperationError);export const isProviderDriverKind = Schema.is(ProviderDriverKind);diff --git a/apps/server/src/orchestration/Layers/provider-command/Turns.ts b/apps/server/src/orchestration/Layers/provider-command/Turns.tsindex bfdb927bd..97c519403 100644--- a/apps/server/src/orchestration/Layers/provider-command/Turns.ts+++ b/apps/server/src/orchestration/Layers/provider-command/Turns.ts@@ -198,7 +198,6 @@ export function createTurns({turnId: null,category: "turn",maximumTokens: AKERU_TURN_USAGE_RESERVATION_TOKENS,- capLimit: respondingBot?.usageCap?.limit ?? Number.MAX_SAFE_INTEGER,provider:[thread.session?.providerName,diff --git a/apps/server/src/orchestration/Layers/snapshot-query/EnvironmentRows.ts b/apps/server/src/orchestration/Layers/snapshot-query/EnvironmentRows.tsindex 2361b655b..a83f2b9d3 100644--- a/apps/server/src/orchestration/Layers/snapshot-query/EnvironmentRows.ts+++ b/apps/server/src/orchestration/Layers/snapshot-query/EnvironmentRows.ts@@ -52,7 +52,7 @@ export function createEnvironmentRows({ sql }: Pick<ProjectionSnapshotDependencibot_id AS "botId", name, title, label, description,disabled_mcp_server_ids_json AS "disabledMcpServerIds", avatar_json AS "avatar",engine_json AS "engine", sandbox, runtime_mode AS "runtimeMode",- usage_cap_json AS "usageCap", image_provider AS "imageProvider", voice_enabled AS "voiceEnabled",+ image_provider AS "imageProvider", voice_enabled AS "voiceEnabled",personality_tone AS "personalityTone",channel_bindings_json AS "channelBindings", group_id AS "groupId",archived_at AS "archivedAt", created_at AS "createdAt", updated_at AS "updatedAt"@@ -127,7 +127,7 @@ export function createEnvironmentRows({ sql }: Pick<ProjectionSnapshotDependencibot_id AS "botId", name, title, label, description,disabled_mcp_server_ids_json AS "disabledMcpServerIds", avatar_json AS "avatar",engine_json AS "engine", sandbox, runtime_mode AS "runtimeMode",- usage_cap_json AS "usageCap", voice_enabled AS "voiceEnabled",+ voice_enabled AS "voiceEnabled",personality_tone AS "personalityTone",channel_bindings_json AS "channelBindings", group_id AS "groupId",archived_at AS "archivedAt", created_at AS "createdAt", updated_at AS "updatedAt"diff --git a/apps/server/src/orchestration/Layers/test-support/OrchestrationRestartSystem.ts b/apps/server/src/orchestration/Layers/test-support/OrchestrationRestartSystem.tsindex 2bd66b6ae..c58773b53 100644--- a/apps/server/src/orchestration/Layers/test-support/OrchestrationRestartSystem.ts+++ b/apps/server/src/orchestration/Layers/test-support/OrchestrationRestartSystem.ts@@ -87,7 +87,6 @@ export async function createRestartSystem() {avatar: { kind: "dither", seed: "restart-bot" },engine: null,sandbox: "local",- usageCap: null,groupId: null,createdAt,});diff --git a/apps/server/src/orchestration/Layers/test-support/ProviderCommandHarness.ts b/apps/server/src/orchestration/Layers/test-support/ProviderCommandHarness.tsindex 602426afe..ab0da5ec3 100644--- a/apps/server/src/orchestration/Layers/test-support/ProviderCommandHarness.ts+++ b/apps/server/src/orchestration/Layers/test-support/ProviderCommandHarness.ts@@ -346,7 +346,6 @@ export function createProviderCommandHarness() {engine: input.botEngine,sandbox: "local",runtimeMode: "approval-required",- usageCap: input.botUsageCap ?? null,groupId: null,createdAt: now,}),@@ -365,7 +364,6 @@ export function createProviderCommandHarness() {engine: input.secondBot.engine,sandbox: "local",runtimeMode: "approval-required",- usageCap: null,groupId: null,createdAt: now,}),diff --git a/apps/server/src/orchestration/Layers/test-support/ProviderCommandMocks.ts b/apps/server/src/orchestration/Layers/test-support/ProviderCommandMocks.tsindex f424f5740..5a9a08e18 100644--- a/apps/server/src/orchestration/Layers/test-support/ProviderCommandMocks.ts+++ b/apps/server/src/orchestration/Layers/test-support/ProviderCommandMocks.ts@@ -59,7 +59,6 @@ export type ProviderCommandHarnessOptions = {readonly engine: { readonly provider: string; readonly model: string };readonly modelSelection?: ModelSelection;};- readonly botUsageCap?: { readonly unit: "tokens"; readonly limit: number } | null;readonly bindTurnFailure?: boolean;readonly unavailableEngine?: boolean | "missing-login";readonly disabledEngine?: boolean;diff --git a/apps/server/src/orchestration/Layers/test-support/RuntimeIngestionHarness.ts b/apps/server/src/orchestration/Layers/test-support/RuntimeIngestionHarness.tsindex fa8e3a4bc..fa4b20387 100644--- a/apps/server/src/orchestration/Layers/test-support/RuntimeIngestionHarness.ts+++ b/apps/server/src/orchestration/Layers/test-support/RuntimeIngestionHarness.ts@@ -201,7 +201,6 @@ export function createRuntimeIngestionHarness() {serverSettings?: Partial<ServerSettings>;threadTitle?: string;botOwned?: boolean;- botUsageCap?: { readonly unit: "tokens"; readonly limit: number } | null;workspaceSubdirectory?: string;}) {const repositoryRoot = makeTempDir("t3-provider-project-");@@ -325,7 +324,6 @@ export function createRuntimeIngestionHarness() {},sandbox: null,runtimeMode: "approval-required",- usageCap: options.botUsageCap ?? null,groupId: null,createdAt,});@@ -489,7 +487,6 @@ export function createRuntimeIngestionHarness() {turnId,category: "turn",maximumTokens: 1_000,- capLimit: 1_000,provider: ProviderDriverKind.make("codex"),model: "gpt-5-codex",createdAt,diff --git a/apps/server/src/orchestration/Layers/test-support/SilenceWatchdogHarness.ts b/apps/server/src/orchestration/Layers/test-support/SilenceWatchdogHarness.tsindex 39392def2..b199a8049 100644--- a/apps/server/src/orchestration/Layers/test-support/SilenceWatchdogHarness.ts+++ b/apps/server/src/orchestration/Layers/test-support/SilenceWatchdogHarness.ts@@ -214,7 +214,6 @@ export function makeSilenceWatchdogHarness() {},sandbox: null,runtimeMode: "approval-required",- usageCap: null,groupId: null,createdAt,});diff --git a/apps/server/src/orchestration/decider.archivedBot.test.ts b/apps/server/src/orchestration/decider.archivedBot.test.tsindex b2962f5ca..b879d87bc 100644--- a/apps/server/src/orchestration/decider.archivedBot.test.ts+++ b/apps/server/src/orchestration/decider.archivedBot.test.ts@@ -38,7 +38,6 @@ function makeBot(archivedAt: string | null = null): OrchestrationBot {engine: null,sandbox: "local",runtimeMode: "full-access",- usageCap: null,imageProvider: null,voiceEnabled: true,channelBindings: [],diff --git a/apps/server/src/orchestration/decider.userInput.test.ts b/apps/server/src/orchestration/decider.userInput.test.tsindex 7564da8fd..0724adff6 100644--- a/apps/server/src/orchestration/decider.userInput.test.ts+++ b/apps/server/src/orchestration/decider.userInput.test.ts@@ -41,7 +41,6 @@ const archivedBot: OrchestrationBot = {imageProvider: null,sandbox: "local",runtimeMode: "full-access",- usageCap: null,voiceEnabled: true,channelBindings: [],groupId: null,diff --git a/apps/server/src/orchestration/decider.voiceTranscript.test.ts b/apps/server/src/orchestration/decider.voiceTranscript.test.tsindex 80bfd8410..d8db50a54 100644--- a/apps/server/src/orchestration/decider.voiceTranscript.test.ts+++ b/apps/server/src/orchestration/decider.voiceTranscript.test.ts@@ -37,7 +37,6 @@ function makeBot(archivedAt: string | null = null): OrchestrationBot {engine: null,sandbox: "local",runtimeMode: "full-access",- usageCap: null,imageProvider: null,voiceEnabled: true,channelBindings: [],diff --git a/apps/server/src/orchestration/decider/bots.ts b/apps/server/src/orchestration/decider/bots.tsindex 9fb97833b..11423a64d 100644--- a/apps/server/src/orchestration/decider/bots.ts+++ b/apps/server/src/orchestration/decider/bots.ts@@ -79,7 +79,6 @@ export const decideBots = Effect.fn("decideBots")(function* ({(command.sandbox === null || command.sandbox === "local"? DEFAULT_LOCAL_EXECUTION_MODE: DEFAULT_RUNTIME_MODE),- usageCap: command.usageCap,imageProvider: command.imageProvider ?? null,personalityTone: command.personalityTone ?? BALANCED_BOT_PERSONALITY_TONE,voiceEnabled: command.voiceEnabled ?? false,@@ -215,7 +214,6 @@ export const decideBots = Effect.fn("decideBots")(function* ({...(command.engine !== undefined ? { engine: command.engine } : {}),...(command.sandbox !== undefined ? { sandbox: command.sandbox } : {}),...(command.runtimeMode !== undefined ? { runtimeMode: command.runtimeMode } : {}),- ...(command.usageCap !== undefined ? { usageCap: command.usageCap } : {}),...(command.imageProvider !== undefined ? { imageProvider: command.imageProvider } : {}),...(command.personalityTone !== undefined? { personalityTone: command.personalityTone }diff --git a/apps/server/src/orchestration/http.ts b/apps/server/src/orchestration/http.tsindex 0a0c78396..424f3da84 100644--- a/apps/server/src/orchestration/http.ts+++ b/apps/server/src/orchestration/http.ts@@ -22,7 +22,6 @@ import * as ProviderRegistry from "../provider/Services/ProviderRegistry.ts";import { deriveProviderInstanceConfigMap } from "../provider/Layers/ProviderInstanceRegistryHydration.ts";import { preflightProvider } from "../provider/providerPreflight.ts";import { SubscriptionAuthService } from "../subscription-auth/service.ts";-import { BotUsageLedger } from "../usage/BotUsageLedger.ts";import { resolveGroupResponderBotId } from "./groupResponder.ts";import { projectThreadDetailSnapshot } from "./ActivityPayloadProjection.ts";import {@@ -58,7 +57,6 @@ export const orchestrationHttpApiLayer = HttpApiBuilder.group(const projectionGroups = yield* ProjectionGroups.ProjectionGroupRepository;const commandReceipts = yield* OrchestrationCommandReceiptRepository;const providerRegistry = yield* Effect.serviceOption(ProviderRegistry.ProviderRegistry);- const botUsageLedger = yield* BotUsageLedger;const config = yield* ServerConfig.ServerConfig;const subscriptionAuth = yield* SubscriptionAuthService.forSecretsDir(config.secretsDir);const serverSettings = yield* Effect.serviceOption(ServerSettings.ServerSettingsService);@@ -387,26 +385,6 @@ export const orchestrationHttpApiLayer = HttpApiBuilder.group(});}}-- if (bot?.usageCap) {- const usage = yield* botUsageLedger- .summarize(bot.botId)- .pipe(- Effect.catch((cause) =>- failEnvironmentInternal("orchestration_dispatch_failed", cause),- ),- );-- if (usage.consumedTokens + usage.reservedTokens >= bot.usageCap.limit) {- yield* cleanupFailedUploadedAttachments(command, normalizedCommand);-- return yield* failEnvironmentInvalidRequest("invalid_command", {- detail: `Usage cap reached for ${bot.name}.`,- unavailability: "usage-cap",- repairAction: "usage",- });- }- }}return yield* dispatchKeepingAcceptedUploads({diff --git a/apps/server/src/orchestration/projector/BotsAndGroups.ts b/apps/server/src/orchestration/projector/BotsAndGroups.tsindex 0590cdb30..49e0607a6 100644--- a/apps/server/src/orchestration/projector/BotsAndGroups.ts+++ b/apps/server/src/orchestration/projector/BotsAndGroups.ts@@ -39,7 +39,6 @@ export function projectBotsAndGroups(engine: payload.engine,sandbox: payload.sandbox,runtimeMode: payload.runtimeMode,- usageCap: payload.usageCap,imageProvider: payload.imageProvider,personalityTone: payload.personalityTone,voiceEnabled: payload.voiceEnabled,@@ -76,7 +75,6 @@ export function projectBotsAndGroups(...(payload.engine !== undefined ? { engine: payload.engine } : {}),...(payload.sandbox !== undefined ? { sandbox: payload.sandbox } : {}),...(payload.runtimeMode !== undefined ? { runtimeMode: payload.runtimeMode } : {}),- ...(payload.usageCap !== undefined ? { usageCap: payload.usageCap } : {}),...(payload.imageProvider !== undefined? { imageProvider: payload.imageProvider }: {}),diff --git a/apps/server/src/orchestration/test-support/BotDeleteFixtures.ts b/apps/server/src/orchestration/test-support/BotDeleteFixtures.tsindex 82a73a2ca..6159752e8 100644--- a/apps/server/src/orchestration/test-support/BotDeleteFixtures.ts+++ b/apps/server/src/orchestration/test-support/BotDeleteFixtures.ts@@ -39,7 +39,6 @@ export function makeBot(input: {engine: null,sandbox: "local",runtimeMode: "full-access",- usageCap: null,imageProvider: null,voiceEnabled: false,channelBindings: [],diff --git a/apps/server/src/orchestration/test-support/DelegationDeciderFixtures.ts b/apps/server/src/orchestration/test-support/DelegationDeciderFixtures.tsindex 6b6246695..7558deee5 100644--- a/apps/server/src/orchestration/test-support/DelegationDeciderFixtures.ts+++ b/apps/server/src/orchestration/test-support/DelegationDeciderFixtures.ts@@ -49,7 +49,6 @@ export function makeBot(id: BotId): OrchestrationBot {engine: null,sandbox: "local",runtimeMode: "approval-required",- usageCap: null,imageProvider: null,voiceEnabled: false,channelBindings: [],diff --git a/apps/server/src/orchestration/test-support/GroupDeciderFixtures.ts b/apps/server/src/orchestration/test-support/GroupDeciderFixtures.tsindex 49448a78e..9dd1da920 100644--- a/apps/server/src/orchestration/test-support/GroupDeciderFixtures.ts+++ b/apps/server/src/orchestration/test-support/GroupDeciderFixtures.ts@@ -73,7 +73,6 @@ export function makeBot(input: {engine: input.provider && input.model ? { provider: input.provider, model: input.model } : null,sandbox: "local",runtimeMode: "full-access",- usageCap: null,imageProvider: null,voiceEnabled: false,channelBindings: [],diff --git a/apps/server/src/persistence/Layers/ProjectionBots.ts b/apps/server/src/persistence/Layers/ProjectionBots.tsindex 16a8ee7e0..44f89fcac 100644--- a/apps/server/src/persistence/Layers/ProjectionBots.ts+++ b/apps/server/src/persistence/Layers/ProjectionBots.ts@@ -1,4 +1,4 @@-import { BotAvatar, BotEngine, BotUsageCap, ChannelBinding, McpServerId } from "@akeru/contracts";+import { BotAvatar, BotEngine, ChannelBinding, McpServerId } from "@akeru/contracts";import * as Effect from "effect/Effect";import * as Layer from "effect/Layer";import * as Option from "effect/Option";@@ -19,7 +19,6 @@ const ProjectionBotDbRow = ProjectionBot.mapFields(Struct.assign({avatar: Schema.fromJsonString(BotAvatar),engine: Schema.NullOr(Schema.fromJsonString(BotEngine)),- usageCap: Schema.NullOr(Schema.fromJsonString(BotUsageCap)),personalityTone: Schema.Number,disabledMcpServerIds: Schema.fromJsonString(Schema.Array(McpServerId)),channelBindings: Schema.fromJsonString(Schema.Array(ChannelBinding)),@@ -39,15 +38,14 @@ const makeProjectionBotRepository = Effect.gen(function* () {execute: (row) => sql`INSERT INTO projection_bots (bot_id, name, title, label, description, disabled_mcp_server_ids_json,- avatar_json, engine_json, sandbox, runtime_mode, usage_cap_json, image_provider,+ avatar_json, engine_json, sandbox, runtime_mode, image_provider,voice_enabled, personality_tone, channel_bindings_json,group_id, archived_at, created_at, updated_at) VALUES (${row.botId}, ${row.name}, ${row.title}, ${row.label}, ${row.description},${JSON.stringify(row.disabledMcpServerIds)}, ${JSON.stringify(row.avatar)},${row.engine === null ? null : JSON.stringify(row.engine)}, ${row.sandbox},- ${row.runtimeMode}, ${row.usageCap === null ? null : JSON.stringify(row.usageCap)},- ${row.imageProvider},+ ${row.runtimeMode}, ${row.imageProvider},${row.voiceEnabled ? 1 : 0}, ${row.personalityTone ?? 50},${JSON.stringify(row.channelBindings ?? [])}, ${row.groupId},${row.archivedAt}, ${row.createdAt},@@ -63,7 +61,6 @@ const makeProjectionBotRepository = Effect.gen(function* () {engine_json = excluded.engine_json,sandbox = excluded.sandbox,runtime_mode = excluded.runtime_mode,- usage_cap_json = excluded.usage_cap_json,image_provider = excluded.image_provider,voice_enabled = excluded.voice_enabled,personality_tone = excluded.personality_tone,@@ -83,7 +80,7 @@ const makeProjectionBotRepository = Effect.gen(function* () {bot_id AS "botId", name, title, label, description,disabled_mcp_server_ids_json AS "disabledMcpServerIds", avatar_json AS "avatar",engine_json AS "engine", sandbox, runtime_mode AS "runtimeMode",- usage_cap_json AS "usageCap", image_provider AS "imageProvider", voice_enabled AS "voiceEnabled",+ image_provider AS "imageProvider", voice_enabled AS "voiceEnabled",personality_tone AS "personalityTone",channel_bindings_json AS "channelBindings", group_id AS "groupId",archived_at AS "archivedAt", created_at AS "createdAt", updated_at AS "updatedAt"@@ -100,7 +97,7 @@ const makeProjectionBotRepository = Effect.gen(function* () {bot_id AS "botId", name, title, label, description,disabled_mcp_server_ids_json AS "disabledMcpServerIds", avatar_json AS "avatar",engine_json AS "engine", sandbox, runtime_mode AS "runtimeMode",- usage_cap_json AS "usageCap", image_provider AS "imageProvider", voice_enabled AS "voiceEnabled",+ image_provider AS "imageProvider", voice_enabled AS "voiceEnabled",personality_tone AS "personalityTone",channel_bindings_json AS "channelBindings", group_id AS "groupId",archived_at AS "archivedAt", created_at AS "createdAt", updated_at AS "updatedAt"diff --git a/apps/server/src/persistence/Services/ProjectionBots.ts b/apps/server/src/persistence/Services/ProjectionBots.tsindex 45145b775..3a98f016f 100644--- a/apps/server/src/persistence/Services/ProjectionBots.ts+++ b/apps/server/src/persistence/Services/ProjectionBots.ts@@ -5,7 +5,6 @@ import {ImageProviderId,BotPersonalityTone,PersistedBotSandbox,- BotUsageCap,ChannelBinding,GroupId,IsoDateTime,@@ -30,7 +29,6 @@ export const ProjectionBot = Schema.Struct({engine: Schema.NullOr(BotEngine),sandbox: PersistedBotSandbox,runtimeMode: RuntimeMode,- usageCap: Schema.NullOr(BotUsageCap),imageProvider: Schema.NullOr(ImageProviderId).pipe(Schema.withDecodingDefault(Effect.succeed(null)),),diff --git a/apps/server/src/portability.test.ts b/apps/server/src/portability.test.tsindex 52f17b00c..cc5c814e8 100644--- a/apps/server/src/portability.test.ts+++ b/apps/server/src/portability.test.ts@@ -115,6 +115,44 @@ describe("portability archive", () => {).toEqual(expect.objectContaining({ botSandboxBrowserSharing: "separate" }));});+ it("imports older archives whose bots still carry the retired token hard stop", () => {+ const snapshot = makeSnapshot();+ const archive = createPortabilityArchive(snapshot, makeSettings(), NOW);++ const legacy = resignArchive(+ archive,+ archive.records.map((record) =>+ record.type === "bot"+ ? { ...record, data: { ...record.data, usageCap: { unit: "tokens", limit: 50_000 } } }+ : record,+ ),+ );++ const parsed = parsePortabilityArchive(JSON.stringify(legacy));++ const commands = commandsForPortabilityImport(+ parsed,+ makeSnapshot(),+ makeSettings(),+ AVAILABLE_PROVIDER_IDS,+ );++ expect(JSON.stringify(commands)).not.toContain("usageCap");+ expect(+ parsed.records.some((record) => record.type === "bot" && "usageCap" in record.data),+ ).toBe(false);+ expect(+ previewPortabilityImport(parsed, makeSnapshot(), makeSettings(), AVAILABLE_PROVIDER_IDS)+ .changes,+ ).toEqual(+ previewPortabilityImport(archive, makeSnapshot(), makeSettings(), AVAILABLE_PROVIDER_IDS)+ .changes,+ );+ expect(+ archive.records.some((record) => record.type === "bot" && "usageCap" in record.data),+ ).toBe(false);+ });+it("removes credentials, local state, Git state, and event internals", () => {const text = serializePortabilityArchive(createPortabilityArchive(makeSnapshot(), makeSettings(), NOW),diff --git a/apps/server/src/portabilityArchive.ts b/apps/server/src/portabilityArchive.tsindex f3e055174..84a6379c6 100644--- a/apps/server/src/portabilityArchive.ts+++ b/apps/server/src/portabilityArchive.ts@@ -113,7 +113,6 @@ export function portableRecords(engine: bot.engine,sandbox: bot.sandbox,runtimeMode: bot.runtimeMode,- usageCap: bot.usageCap,imageProvider: bot.imageProvider,personalityTone: bot.personalityTone ?? BALANCED_BOT_PERSONALITY_TONE,voiceEnabled: bot.voiceEnabled,diff --git a/apps/server/src/portabilityCommands.ts b/apps/server/src/portabilityCommands.tsindex cb111fa16..20a5285ce 100644--- a/apps/server/src/portabilityCommands.ts+++ b/apps/server/src/portabilityCommands.ts@@ -350,7 +350,6 @@ export function commandsForPortabilityImport(engine: record.data.engine,sandbox: record.data.sandbox,runtimeMode: record.data.runtimeMode,- usageCap: record.data.usageCap,imageProvider: record.data.imageProvider,personalityTone: record.data.personalityTone ?? BALANCED_BOT_PERSONALITY_TONE,voiceEnabled: record.data.voiceEnabled,diff --git a/apps/server/src/portabilityParse.ts b/apps/server/src/portabilityParse.tsindex ebf42b1fc..173a1ccba 100644--- a/apps/server/src/portabilityParse.ts+++ b/apps/server/src/portabilityParse.ts@@ -271,5 +271,15 @@ export function parsePortabilityArchive(contents: string) {}}- return archive;+ // Older archives sign the retired token hard stop; once verified, drop it so+ // previews and imports compare only the settings bots still have.+ return {+ ...archive,+ records: archive.records.map((record) => {+ if (record.type !== "bot" || !("usageCap" in record.data)) return record;+ const { usageCap: _usageCap, ...data } = record.data;++ return { ...record, data };+ }),+ };}diff --git a/apps/server/src/portabilityTestSupport.ts b/apps/server/src/portabilityTestSupport.tsindex 67dbd344d..40e1bbb33 100644--- a/apps/server/src/portabilityTestSupport.ts+++ b/apps/server/src/portabilityTestSupport.ts@@ -139,7 +139,6 @@ export function makeSnapshot(engine: { provider: "codex", model: "gpt-5.6-sol" },sandbox: "local",runtimeMode: "full-access",- usageCap: null,imageProvider: null,voiceEnabled: true,channelBindings: [],diff --git a/apps/server/src/provider/AkeruBotStateRuntime.test.ts b/apps/server/src/provider/AkeruBotStateRuntime.test.tsindex 6d12f3a98..6a667463c 100644--- a/apps/server/src/provider/AkeruBotStateRuntime.test.ts+++ b/apps/server/src/provider/AkeruBotStateRuntime.test.ts@@ -25,7 +25,6 @@ function snapshot(archivedAt: string | null = null) {engine: null,sandbox: "local" as const,runtimeMode: "approval-required" as const,- usageCap: null,imageProvider: null,voiceEnabled: false,channelBindings: [],@@ -45,7 +44,6 @@ function snapshot(archivedAt: string | null = null) {engine: null,sandbox: "local" as const,runtimeMode: "approval-required" as const,- usageCap: null,imageProvider: null,voiceEnabled: false,channelBindings: [],diff --git a/apps/server/src/provider/AkeruChannelRuntime.test.ts b/apps/server/src/provider/AkeruChannelRuntime.test.tsindex f1aeae388..dc00519d4 100644--- a/apps/server/src/provider/AkeruChannelRuntime.test.ts+++ b/apps/server/src/provider/AkeruChannelRuntime.test.ts@@ -34,7 +34,6 @@ const bot = (id: BotId, groupId: GroupId | null) => ({engine: null,sandbox: "local" as const,runtimeMode: "approval-required" as const,- usageCap: null,imageProvider: null,voiceEnabled: false,channelBindings: [],diff --git a/apps/server/src/provider/AkeruOpenAIProvider.ts b/apps/server/src/provider/AkeruOpenAIProvider.tsindex ce6b9cdbd..6886043e7 100644--- a/apps/server/src/provider/AkeruOpenAIProvider.ts+++ b/apps/server/src/provider/AkeruOpenAIProvider.ts@@ -1,3 +1,8 @@+import {+ buildOpenAICodexOAuthFetch,+ createCodexMiddleware,+} from "@mastra/code-sdk/providers/openai-codex";+import type { AuthStorage } from "@mastra/code-sdk/auth/storage";import { createOpenAI } from "@ai-sdk/openai";import type { ApiKeyCredential } from "../subscription-auth/types.ts";import { subscriptionRequestUrl } from "../subscription-auth/runtime.ts";@@ -41,3 +46,37 @@ export function akeruOpenAIProvider(>,}).responses(modelId);}++/** Preserves the SDK's OAuth default while accepting raw native efforts without coercion. */+export function akeruCodexOAuthProvider(modelId: string, authStorage: AuthStorage) {+ const model = createOpenAI({+ apiKey: "oauth-placeholder",+ fetch: buildOpenAICodexOAuthFetch({ authStorage }),+ }).responses(modelId);++ const middleware = createCodexMiddleware();+ const defaultMiddleware = createCodexMiddleware("medium");++ const transform = (params: Parameters<typeof model.doGenerate>[0], type: "generate" | "stream") =>+ (params.providerOptions?.openai?.reasoningEffort === undefined+ ? defaultMiddleware+ : middleware+ ).transformParams?.({ type, params, model });++ return {+ specificationVersion: model.specificationVersion,+ provider: model.provider,+ modelId: model.modelId,+ supportedUrls: model.supportedUrls,+ async doGenerate(params: Parameters<typeof model.doGenerate>[0]) {+ const transformed = await transform(params, "generate");++ return model.doGenerate(transformed ?? params);+ },+ async doStream(params: Parameters<typeof model.doStream>[0]) {+ const transformed = await transform(params, "stream");++ return model.doStream(transformed ?? params);+ },+ };+}diff --git a/apps/server/src/provider/ClaudeReasoningCapabilities.ts b/apps/server/src/provider/ClaudeReasoningCapabilities.tsnew file mode 100644index 000000000..b119ae591--- /dev/null+++ b/apps/server/src/provider/ClaudeReasoningCapabilities.ts@@ -0,0 +1,38 @@+import { createModelCapabilities } from "@akeru/shared/model";+import { ProviderDriverKind } from "@akeru/contracts";+import { CLAUDE_MODEL_CATALOG, getClaudeModelCapabilities } from "./Layers/claude/ClaudeModels.ts";+import { reasoningCapabilities } from "./ReasoningOptions.ts";+import { BUNDLED_MODEL_CATALOG, catalogModelsFor, type CatalogModel } from "./modelCatalogData.ts";++/** The subscription transport advertises native controls, without CLI workflow choices. */+export function claudeHarnessCapabilities(slug: string, entry: CatalogModel | undefined) {+ const previous = getClaudeModelCapabilities(slug);++ const builtIn = CLAUDE_MODEL_CATALOG.find((model) => model.slug === slug);++ const builtInEffort = builtIn?.capabilities?.optionDescriptors?.find(+ (descriptor) => descriptor.id === "effort" && descriptor.type === "select",+ );++ const fallback =+ builtInEffort?.type === "select"+ ? { id: slug, name: slug, efforts: builtInEffort.options.map((option) => option.id) }+ : undefined;++ const driver = ProviderDriverKind.make("claudeAgent");++ const bundled = catalogModelsFor(BUNDLED_MODEL_CATALOG, driver).find(+ (model) => model.id === slug,+ );++ const native = reasoningCapabilities(driver, slug, entry ?? bundled ?? fallback);++ return createModelCapabilities({+ optionDescriptors: [+ ...(native.optionDescriptors ?? []),+ ...(previous.optionDescriptors ?? []).filter(+ (descriptor) => descriptor.id === "contextWindow" || descriptor.id === "thinking",+ ),+ ],+ });+}diff --git a/apps/server/src/provider/Drivers/HarnessProviderDrivers.test.ts b/apps/server/src/provider/Drivers/HarnessProviderDrivers.test.tsindex 8b6ae029c..5107b8a28 100644--- a/apps/server/src/provider/Drivers/HarnessProviderDrivers.test.ts+++ b/apps/server/src/provider/Drivers/HarnessProviderDrivers.test.ts@@ -424,7 +424,7 @@ it.layer(testLayer)("Harness provider drivers without CLIs", (it) => {?.optionDescriptors,).toEqual(expect.arrayContaining([- expect.objectContaining({ id: "reasoningEffort", currentValue: "medium" }),+ expect.objectContaining({ id: "reasoningEffort" }),expect.objectContaining({ id: "serviceTier", currentValue: "default" }),]),);diff --git a/apps/server/src/provider/HarnessProviderStatus.test.ts b/apps/server/src/provider/HarnessProviderStatus.test.tsindex 7185d3b81..85615f783 100644--- a/apps/server/src/provider/HarnessProviderStatus.test.ts+++ b/apps/server/src/provider/HarnessProviderStatus.test.ts@@ -44,26 +44,26 @@ describe("codexModelCapabilities", () => {options("gpt-next", {id: "gpt-next",name: "GPT Next",- efforts: ["none", "low", "medium", "xhigh"],+ efforts: ["default", "none", "low", "medium", "xhigh"],fast: true,}),).toEqual([- ["reasoningEffort", ["off", "low", "medium", "xhigh"]],+ ["reasoningEffort", ["default", "none", "low", "medium", "xhigh"]],["serviceTier", ["default", "priority"]],]);expect(options("gpt-slow", { id: "gpt-slow", name: "GPT Slow", efforts: ["high"] })).toEqual([- ["reasoningEffort", ["high"]],+ ["reasoningEffort", ["default", "high"]],]);});- it("drops Max where the Codex transport would send Extra High instead", () => {+ it("retains catalog levels without the SDK thinking-level coercion", () => {const efforts = ["low", "medium", "high", "xhigh", "max"];expect(options("gpt-daybreak-blue-latest", { id: "x", name: "X", efforts })).toEqual([- ["reasoningEffort", ["low", "medium", "high", "xhigh"]],+ ["reasoningEffort", ["default", ...efforts]],]);expect(options("gpt-6.1-sol", { id: "x", name: "X", efforts })).toEqual([- ["reasoningEffort", efforts],+ ["reasoningEffort", ["default", ...efforts]],]);});});diff --git a/apps/server/src/provider/HarnessProviderStatus.ts b/apps/server/src/provider/HarnessProviderStatus.tsindex b2e44d749..a5d5072b9 100644--- a/apps/server/src/provider/HarnessProviderStatus.ts+++ b/apps/server/src/provider/HarnessProviderStatus.ts@@ -15,14 +15,10 @@ import {} from "../subscription-auth/service.ts";import type { ProviderInstance } from "./ProviderDriver.ts";import * as ModelCatalog from "./ModelCatalog.ts";-import {- BUNDLED_MODEL_CATALOG,- catalogModelsFor,- harnessEffortLevels,- type CatalogModel,-} from "./modelCatalogData.ts";+import { BUNDLED_MODEL_CATALOG, catalogModelsFor, type CatalogModel } from "./modelCatalogData.ts";import type { ServerProviderDraft } from "./providerSnapshot.ts";-import { getClaudeModelCapabilities } from "./Layers/ClaudeProvider.ts";+import { claudeHarnessCapabilities } from "./ClaudeReasoningCapabilities.ts";+import { reasoningCapabilities } from "./ReasoningOptions.ts";export const GROK_HARNESS_MODELS = ["grok-4.6",@@ -93,19 +89,19 @@ const effortLabel = (level: string) =>/*** Codex option descriptors. Catalog models offer exactly the efforts and* service tiers models.dev lists; slugs outside the catalog fall back to- * Mastra's per-model thinking levels and always offer Fast.+ * Mastra's known per-model levels and offer Fast.*/export function codexModelCapabilities(slug: string, entry: CatalogModel | undefined) {- // The Codex transport downgrades `max` to `xhigh` for models the harness- // SDK does not know support it, so only offer levels it will actually send.const supported: ReadonlyArray<string> = getAvailableThinkingLevelsForModel(`openai/${slug}`);const levels =- entry?.efforts !== undefined- ? harnessEffortLevels(entry.efforts).filter((level) => supported.includes(level))- : supported;+ entry !== undefined+ ? (entry.efforts ?? []).filter((level) =>+ ["none", "minimal", "low", "medium", "high", "xhigh", "max"].includes(level),+ )+ : supported.map((level) => (level === "off" ? "none" : level));- const defaultLevel = levels.includes("medium") ? "medium" : levels[0];+ const defaultLevel = levels[0];return createModelCapabilities({optionDescriptors: [@@ -116,12 +112,14 @@ export function codexModelCapabilities(slug: string, entry: CatalogModel | undefid: "reasoningEffort",label: "Reasoning",type: "select" as const,- currentValue: defaultLevel,- options: levels.map((level) => ({- id: level,- label: effortLabel(level),- ...(level === defaultLevel ? { isDefault: true } : {}),- })),+ currentValue: "default",+ options: [+ { id: "default", label: "Provider default", isDefault: true },+ ...levels.map((level) => ({+ id: level,+ label: effortLabel(level),+ })),+ ],},]),...(entry === undefined || entry.fast === true@@ -223,7 +221,21 @@ export const makeHarnessProviderStatus = Effect.fn("makeHarnessProviderStatus")(.map((model) => {const name = model.isCustom ? undefined : entries.get(model.slug)?.name;- return name ? { ...model, name } : model;+ return {+ ...model,+ ...(name ? { name } : {}),+ capabilities: model.isCustom+ ? model.capabilities+ : input.driver === "codex"+ ? codexModelCapabilities(model.slug, entries.get(model.slug))+ : input.driver === "claudeAgent"+ ? claudeHarnessCapabilities(model.slug, entries.get(model.slug))+ : reasoningCapabilities(+ input.driver,+ model.slug === "grok-build" ? "grok-4.6" : model.slug,+ entries.get(model.slug === "grok-build" ? "grok-4.6" : model.slug),+ ),+ };}),...modelIds.filter((slug) => !draft.models.some((model) => !model.isCustom && model.slug === slug))@@ -236,8 +248,8 @@ export const makeHarnessProviderStatus = Effect.fn("makeHarnessProviderStatus")(: {}),capabilities: Match.value(input.driver).pipe(Match.when("codex", () => codexModelCapabilities(slug, entries.get(slug))),- Match.when("claudeAgent", () => getClaudeModelCapabilities(slug)),- Match.orElse(() => null),+ Match.when("claudeAgent", () => claudeHarnessCapabilities(slug, entries.get(slug))),+ Match.orElse(() => reasoningCapabilities(input.driver, slug, entries.get(slug))),),})),];diff --git a/apps/server/src/provider/Layers/AgentController.memoryCadence.admission.test.ts b/apps/server/src/provider/Layers/AgentController.memoryCadence.admission.test.tsindex b7ab12a90..8211391ff 100644--- a/apps/server/src/provider/Layers/AgentController.memoryCadence.admission.test.ts+++ b/apps/server/src/provider/Layers/AgentController.memoryCadence.admission.test.ts@@ -27,7 +27,7 @@ import { createBotMemoryToolHandler } from "../../memory/BotMemoryToolHandlers.timport * as McpMemoryToolSession from "../../mcp/McpMemoryToolSession.ts";import { AgentController } from "../Services/AgentController.ts";import * as ProjectionSnapshotQuery from "../../orchestration/Services/ProjectionSnapshotQuery.ts";-import { BotUsageCapExceeded } from "../../usage/BotUsageLedger.ts";+import { PersistenceSqlError } from "../../persistence/Errors.ts";import {codexThreadId,claudeThreadId,@@ -410,7 +410,7 @@ describe("AgentControllerLive", () => {yield* controller.sendTurn({threadId: codexThreadId,input: "Remember this.",- botUsage: { botId, capLimit: 50_000 },+ botUsage: { botId },});const options = mastra.harnessOptions[0]!;@@ -426,7 +426,6 @@ describe("AgentControllerLive", () => {threadId: codexThreadId,category: "observer",maximumTokens: 32_000,- capLimit: 50_000,provider: "codex",model: "gpt-5.6-sol",}),@@ -449,20 +448,12 @@ describe("AgentControllerLive", () => {});usageLedger.reserve.mockImplementationOnce(() =>- Effect.fail(- new BotUsageCapExceeded({- botId,- limit: 50_000,- consumedTokens: 20_000,- reservedTokens: 30_000,- requestedTokens: 32_000,- }),- ),+ Effect.fail(new PersistenceSqlError({ operation: "BotUsageLedger.reserve" })),);yield* Effect.promise(() =>expect(options.startMemoryCall!({ threadId: codexThreadId, category: "reflector" }),- ).rejects.toBeInstanceOf(BotUsageCapExceeded),+ ).rejects.toBeInstanceOf(PersistenceSqlError),);mastra.finishSend();}),diff --git a/apps/server/src/provider/Layers/AgentController.reasoning.test.ts b/apps/server/src/provider/Layers/AgentController.reasoning.test.tsnew file mode 100644index 000000000..1bf2763ff--- /dev/null+++ b/apps/server/src/provider/Layers/AgentController.reasoning.test.ts@@ -0,0 +1,222 @@+import { it } from "@effect/vitest";+import * as Effect from "effect/Effect";+import { describe, expect, vi } from "vite-plus/test";+import { ProviderDriverKind, type ModelSelection } from "@akeru/contracts";+import type { AkeruMastraState } from "../mastra/AkeruHarnessTypes.ts";+import { reasoningCapabilities } from "../ReasoningOptions.ts";+import { claudeHarnessCapabilities } from "../ClaudeReasoningCapabilities.ts";+import { AgentController } from "../Services/AgentController.ts";+import {+ codexInstanceId,+ codexThreadId,+ codexSelection,+} from "./test-support/agentControllerFixtures.ts";+import {+ makeBridge,+ makeInstanceSnapshot,+ provideController,+} from "./test-support/agentControllerLayers.ts";+import { mastraHarnessFixture } from "./test-support/agentControllerHarness.ts";++const codexStart = {+ threadId: codexThreadId,+ provider: ProviderDriverKind.make("codex"),+ providerInstanceId: codexInstanceId,+ runtimeMode: "approval-required" as const,+};++function reasoningBridge(status: "ready" | "warning", legacy = false, fixed = false) {+ const bridge = makeBridge();++ return {+ ...bridge.service,+ getInstanceInfo: (instanceId: ModelSelection["instanceId"]) =>+ bridge.service.getInstanceInfo(instanceId).pipe(+ Effect.map((info) => ({+ ...info,+ instanceSnapshot: {+ ...makeInstanceSnapshot(instanceId, info.driverKind, {+ models: ["gpt-5.6-sol"],+ status,+ }),+ models: [+ {+ slug: "gpt-5.6-sol",+ name: "Sol",+ isCustom: false,+ capabilities: legacy+ ? null+ : fixed+ ? { optionDescriptors: [] }+ : reasoningCapabilities(info.driverKind, "gpt-5.6-sol", {+ id: "gpt-5.6-sol",+ name: "Sol",+ efforts: ["none", "low", "high"],+ }),+ },+ ],+ },+ })),+ ),+ };+}++const resolve = (+ controller: AgentController["Service"],+ model: string,+ options?: ModelSelection["options"],+ provider = "codex",+) =>+ controller.resolveEngine({+ threadId: codexThreadId,+ engine: { provider, model, ...(options ? { options } : {}) },+ fallback: codexSelection,+ mode: "default",+ botConversation: true,+ });++describe("effective bot reasoning at the controller boundary", () => {+ it.effect(+ "rejects invalid bot override options from a settled snapshot and preserves unrelated option ids",+ () => {+ const mastra = mastraHarnessFixture();+ const bridge = reasoningBridge("ready");++ return provideController(+ Effect.gen(function* () {+ const controller = yield* AgentController;++ const failure = yield* Effect.flip(+ resolve(controller, "gpt-5.6-sol", [{ id: "reasoningEffort", value: "medium" }]),+ );++ expect(failure._tag).toBe("ProviderValidationError");++ const valid = yield* resolve(controller, "gpt-5.6-sol", [+ { id: "reasoningEffort", value: "off" },+ { id: "futureOption", value: "custom" },+ ]);++ expect(valid.modelSelection.options).toEqual([+ { id: "reasoningEffort", value: "off" },+ { id: "futureOption", value: "custom" },+ ]);+ }),+ bridge,+ mastra.factory,+ );+ },+ );++ it.effect("rejects stale reasoning on a settled fixed model after the bot override", () => {+ const mastra = mastraHarnessFixture();++ return provideController(+ Effect.gen(function* () {+ const controller = yield* AgentController;++ const failure = yield* Effect.flip(+ resolve(controller, "gpt-5.6-sol", [{ id: "reasoningEffort", value: "high" }]),+ );++ expect(failure._tag).toBe("ProviderValidationError");+ expect(mastra.createSession).not.toHaveBeenCalled();++ const reset = yield* resolve(controller, "gpt-5.6-sol", [+ { id: "reasoningEffort", value: "default" },+ ]);++ expect(reset.modelSelection.model).toBe("gpt-5.6-sol");+ }),+ reasoningBridge("ready", false, true),+ mastra.factory,+ );+ });++ it.effect.each([+ { status: "warning", legacy: false },+ { status: "ready", legacy: true },+ ] as const)("handles $status / legacy=$legacy snapshots gracefully", ({ status, legacy }) => {+ const mastra = mastraHarnessFixture();++ return provideController(+ Effect.gen(function* () {+ const controller = yield* AgentController;++ const resolved = yield* resolve(controller, "gpt-5.6-sol", [+ { id: "reasoningEffort", value: "medium" },+ ]);++ expect(resolved.modelSelection.options).toEqual([+ { id: "reasoningEffort", value: "medium" },+ ]);+ }),+ reasoningBridge(status, legacy),+ mastra.factory,+ );+ });++ it.effect(+ "replaces provider options in merged state and clears them on reset and session reuse",+ () => {+ const mastra = mastraHarnessFixture();+ const bridge = makeBridge();+ let state: AkeruMastraState = {};+ mastra.session.state.get = () => state;+ vi.spyOn(mastra.session.state, "set").mockImplementation(async (updates) => {+ state = { ...state, ...updates };+ });++ return provideController(+ Effect.gen(function* () {+ const controller = yield* AgentController;+ yield* resolve(controller, "gpt-5.6-sol", [+ { id: "reasoningEffort", value: "high" },+ { id: "serviceTier", value: "priority" },+ ]);+ yield* controller.startSession(codexThreadId, codexStart);++ expect(state.modelOptions).toEqual({ reasoningEffort: "high", serviceTier: "priority" });+ yield* resolve(+ controller,+ "claude-opus-5-5",+ [{ id: "effort", value: "low" }],+ "claudeAgent",+ );++ expect(state.modelOptions).toEqual({+ namespace: "anthropic",+ effort: "low",+ thinking: true,+ thinkingMode: "adaptive",+ });+ yield* resolve(+ controller,+ "claude-haiku-4-5",+ [{ id: "thinking", value: false }],+ "claudeAgent",+ );++ expect(state.modelOptions).toEqual({+ namespace: "anthropic",+ thinking: false,+ thinkingMode: "enabled",+ });+ yield* resolve(controller, "gpt-5.6-sol", [{ id: "reasoningEffort", value: "default" }]);+ yield* controller.startSession(codexThreadId, codexStart);++ expect(state.modelOptions).toEqual({});+ expect(+ claudeHarnessCapabilities("claude-opus-4-5", {+ id: "claude-opus-4-5",+ name: "Opus",+ efforts: ["low", "medium", "high"],+ }).optionDescriptors?.find((option) => option.id === "thinking"),+ ).toBeUndefined();+ }),+ bridge.service,+ mastra.factory,+ );+ },+ );+});diff --git a/apps/server/src/provider/Layers/OpenCodeAdapter.test.ts b/apps/server/src/provider/Layers/OpenCodeAdapter.test.tsindex cc3ee4dd1..44457d2f0 100644--- a/apps/server/src/provider/Layers/OpenCodeAdapter.test.ts+++ b/apps/server/src/provider/Layers/OpenCodeAdapter.test.ts@@ -27,6 +27,7 @@ import {openCodeAdapterTestSettings,makeOpenCodeAdapterHarness,} from "./test-support/openCodeAdapterHarness.ts";+import { OPENCODE_PROVIDER_DEFAULT_VARIANT } from "./opencode/OpenCodeProtocol.ts";const { runtimeMock, OpenCodeRuntimeTestDouble, OpenCodeAdapterTestLayer } =makeOpenCodeAdapterHarness();@@ -452,6 +453,39 @@ it.layer(OpenCodeAdapterTestLayer)("OpenCodeAdapterLive", (it) => {variant: "high",parts: [{ type: "text", text: "Fix it" }],});++ yield* adapter.sendTurn({+ threadId: asThreadId("thread-custom-instance"),+ input: "Again",+ modelSelection: createModelSelection(+ ProviderInstanceId.make("opencode_zen"),+ "anthropic/claude-sonnet-4-5",+ [{ id: "variant", value: OPENCODE_PROVIDER_DEFAULT_VARIANT }],+ ),+ });++ NodeAssert.deepEqual(runtimeMock.state.promptCalls.at(-1), {+ sessionID: "http://127.0.0.1:9999/session",+ model: { providerID: "anthropic", modelID: "claude-sonnet-4-5" },+ parts: [{ type: "text", text: "Again" }],+ });++ yield* adapter.sendTurn({+ threadId: asThreadId("thread-custom-instance"),+ input: "Once more",+ modelSelection: createModelSelection(+ ProviderInstanceId.make("opencode_zen"),+ "anthropic/claude-sonnet-4-5",+ [{ id: "variant", value: "default" }],+ ),+ });++ NodeAssert.deepEqual(runtimeMock.state.promptCalls.at(-1), {+ sessionID: "http://127.0.0.1:9999/session",+ model: { providerID: "anthropic", modelID: "claude-sonnet-4-5" },+ variant: "default",+ parts: [{ type: "text", text: "Once more" }],+ });}).pipe(Effect.provide(adapterLayer));});});diff --git a/apps/server/src/provider/Layers/OpenCodeAdapter.ts b/apps/server/src/provider/Layers/OpenCodeAdapter.tsindex d536dc9a8..aec97f0e4 100644--- a/apps/server/src/provider/Layers/OpenCodeAdapter.ts+++ b/apps/server/src/provider/Layers/OpenCodeAdapter.ts@@ -58,6 +58,7 @@ import {isOpenCodeNotFound,isSameOpenCodeDirectory,nowIso,+ OPENCODE_PROVIDER_DEFAULT_VARIANT,toRequestError,ensureSessionContext,updateProviderSession,@@ -319,7 +320,8 @@ export function makeOpenCodeAdapter(}const agent = getModelSelectionStringOptionValue(modelSelection, "agent");- const variant = getModelSelectionStringOptionValue(modelSelection, "variant");+ const savedVariant = getModelSelectionStringOptionValue(modelSelection, "variant");+ const variant = savedVariant === OPENCODE_PROVIDER_DEFAULT_VARIANT ? undefined : savedVariant;context.activeTurnId = turnId;context.activeAgent = agent;diff --git a/apps/server/src/provider/Layers/OpenCodeProvider.test.ts b/apps/server/src/provider/Layers/OpenCodeProvider.test.tsindex 98801df65..9b6bc965c 100644--- a/apps/server/src/provider/Layers/OpenCodeProvider.test.ts+++ b/apps/server/src/provider/Layers/OpenCodeProvider.test.ts@@ -18,6 +18,7 @@ import {type OpenCodeRuntimeShape,} from "../opencodeRuntime.ts";import { checkOpenCodeProviderStatus } from "./OpenCodeProvider.ts";+import { OPENCODE_PROVIDER_DEFAULT_VARIANT } from "./opencode/OpenCodeProtocol.ts";import type { OpenCodeInventory } from "../opencodeRuntime.ts";const decodeOpenCodeSettings = Schema.decodeSync(OpenCodeSettings);@@ -194,7 +195,7 @@ it.layer(testLayer)("checkOpenCodeProviderStatus", (it) => {}).pipe(Effect.provide(TestClock.layer())),);- it.effect("emits OpenCode variant defaults so trait picker can resolve a visible selection", () =>+ it.effect("offers Provider default ahead of every OpenCode variant", () =>Effect.gen(function* () {runtimeMock.state.inventory = {providerList: {@@ -211,6 +212,7 @@ it.layer(testLayer)("checkOpenCodeProviderStatus", (it) => {id: "gpt-5.4",name: "GPT-5.4",variants: {+ default: {},none: {},low: {},medium: {},@@ -240,9 +242,18 @@ it.layer(testLayer)("checkOpenCodeProviderStatus", (it) => {);NodeAssert.ok(variantDescriptor && variantDescriptor.type === "select");- NodeAssert.equal(- variantDescriptor.options.find((option) => option.isDefault === true)?.id,- "medium",+ NodeAssert.equal(variantDescriptor.currentValue, OPENCODE_PROVIDER_DEFAULT_VARIANT);+ NodeAssert.deepEqual(+ variantDescriptor.options.map((option) => [option.id, option.isDefault === true]),+ [+ [OPENCODE_PROVIDER_DEFAULT_VARIANT, true],+ ["default", false],+ ["none", false],+ ["low", false],+ ["medium", false],+ ["high", false],+ ["xhigh", false],+ ],);const agentDescriptor = model.capabilities?.optionDescriptors?.find(diff --git a/apps/server/src/provider/Layers/OpenCodeProvider.ts b/apps/server/src/provider/Layers/OpenCodeProvider.tsindex 36ec4f324..7c35900bd 100644--- a/apps/server/src/provider/Layers/OpenCodeProvider.ts+++ b/apps/server/src/provider/Layers/OpenCodeProvider.ts@@ -25,6 +25,7 @@ import {type OpenCodeInventory,} from "../opencodeRuntime.ts";import type { Agent } from "@opencode-ai/sdk/v2";+import { OPENCODE_PROVIDER_DEFAULT_VARIANT } from "./opencode/OpenCodeProtocol.ts";const OPENCODE_PRESENTATION = {displayName: "OpenCode",@@ -154,25 +155,6 @@ function titleCaseSlug(value: string): string {return segments.join(" ");}-function inferDefaultVariant(- providerID: string,- variants: ReadonlyArray<string>,-): string | undefined {- if (variants.length === 1) {- return variants[0];- }-- if (providerID === "anthropic" || providerID.startsWith("google")) {- return variants.includes("high") ? "high" : undefined;- }-- if (providerID === "openai" || providerID === "opencode") {- return variants.includes("medium") ? "medium" : variants.includes("high") ? "high" : undefined;- }-- return undefined;-}-function inferDefaultAgent(agents: ReadonlyArray<Agent>): string | undefined {return agents.find((agent) => agent.name === "build")?.name ?? agents[0]?.name ?? undefined;}@@ -182,18 +164,17 @@ const DEFAULT_OPENCODE_MODEL_CAPABILITIES: ModelCapabilities = createModelCapabi});function openCodeCapabilitiesForModel(input: {- readonly providerID: string;readonly model: OpenCodeInventory["providerList"]["all"][number]["models"][string];readonly agents: ReadonlyArray<Agent>;}): ModelCapabilities {const variantValues = Object.keys(input.model.variants ?? {});- const defaultVariant = inferDefaultVariant(input.providerID, variantValues);- const variantOptions = variantValues.map((value) =>- defaultVariant === value- ? { id: value, label: titleCaseSlug(value), isDefault: true as const }- : { id: value, label: titleCaseSlug(value) },- );+ // OpenCode picks its own default variant from user config, so Akeru offers+ // that choice instead of guessing it; an explicit level is always forwarded.+ const variantOptions = [+ { id: OPENCODE_PROVIDER_DEFAULT_VARIANT, label: "Provider default", isDefault: true as const },+ ...variantValues.map((value) => ({ id: value, label: titleCaseSlug(value) })),+ ];const primaryAgents = input.agents.filter((agent) => !agent.hidden && (agent.mode === "primary" || agent.mode === "all"),@@ -209,14 +190,14 @@ function openCodeCapabilitiesForModel(input: {return createModelCapabilities({optionDescriptors: [- ...(variantOptions.length > 0+ ...(variantValues.length > 0? [{id: "variant",label: "Variant",type: "select" as const,options: variantOptions,- ...(defaultVariant ? { currentValue: defaultVariant } : {}),+ currentValue: OPENCODE_PROVIDER_DEFAULT_VARIANT,},]: []),@@ -258,7 +239,6 @@ function flattenOpenCodeModels(input: OpenCodeInventory): ReadonlyArray<ServerPr...(subProvider ? { subProvider } : {}),isCustom: false,capabilities: openCodeCapabilitiesForModel({- providerID: provider.id,model,agents: input.agents,}),diff --git a/apps/server/src/provider/Layers/agentController/ControllerState.ts b/apps/server/src/provider/Layers/agentController/ControllerState.tsindex 6c457c6b5..a98dba6bf 100644--- a/apps/server/src/provider/Layers/agentController/ControllerState.ts+++ b/apps/server/src/provider/Layers/agentController/ControllerState.ts@@ -110,10 +110,7 @@ export const createControllerState = Effect.fn("createControllerState")(function}>();- const memoryUsageByThread = new Map<- string,- { readonly botId: BotId; readonly capLimit: number; turnId: TurnId }- >();+ const memoryUsageByThread = new Map<string, { readonly botId: BotId; turnId: TurnId }>();const { clearPreviewMcpSession, preparePreviewMcpSession } = createPreviewMcpSessions({options,diff --git a/apps/server/src/provider/Layers/agentController/Conversation.ts b/apps/server/src/provider/Layers/agentController/Conversation.tsindex 5efbd06e8..e0361916d 100644--- a/apps/server/src/provider/Layers/agentController/Conversation.ts+++ b/apps/server/src/provider/Layers/agentController/Conversation.ts@@ -74,10 +74,7 @@ export function createConversation(deps: {reason?: string,) => void;readonly bundle: AkeruControllerHarness;- readonly memoryUsageByThread: Map<- string,- { readonly botId: BotId; readonly capLimit: number; turnId: TurnId }- >;+ readonly memoryUsageByThread: Map<string, { readonly botId: BotId; turnId: TurnId }>;readonly workerRuntime: AkeruWorkerRuntime;readonly workerTurnDefaults: Map<string,diff --git a/apps/server/src/provider/Layers/agentController/EngineRouting.ts b/apps/server/src/provider/Layers/agentController/EngineRouting.tsindex 765d49f03..6a4e25750 100644--- a/apps/server/src/provider/Layers/agentController/EngineRouting.ts+++ b/apps/server/src/provider/Layers/agentController/EngineRouting.ts@@ -1,3 +1,4 @@+import { invalidReasoningSelection, type AkeruModelOptions } from "../../ReasoningOptions.ts";import { resolveClaudeApiModelId } from "../ClaudeProvider.ts";import { ProviderDriverKind } from "@akeru/contracts";import { ProviderInstanceId } from "@akeru/contracts";@@ -39,9 +40,7 @@ export function createEngineRouting(deps: {readonly mutationLock: Semaphore.Semaphore;readonly resolvedByThread: Map<string, ResolvedEngine>;readonly sessions: Map<string, ActiveSession>;- readonly mastraModelOptions: (- resolved: ResolvedEngine,- ) => { serviceTier?: string; reasoningEffort?: string } | undefined;+ readonly mastraModelOptions: (resolved: ResolvedEngine) => AkeruModelOptions | undefined;readonly runMastra: <A>(operation: string,run: (signal: AbortSignal) => Promise<A>,@@ -87,7 +86,21 @@ export function createEngineRouting(deps: {) {const advertised = routing.instanceSnapshot.models;- if (advertised.length > 0 && !advertised.some((entry) => entry.slug === model)) {+ const advertisedModel = advertised.find((entry) => entry.slug === model);++ const invalidOption = invalidReasoningSelection(+ modelSelection,+ advertisedModel?.capabilities,+ );++ if (invalidOption) {+ return yield* new ProviderValidationError({+ operation: "AgentController.inspectEngine",+ issue: invalidOption,+ });+ }++ if (advertised.length > 0 && !advertisedModel) {const name = routing.instanceSnapshot.displayName ?? routing.driverKind;return yield* new AgentControllerUnsupportedEngineError({@@ -182,7 +195,7 @@ export function createEngineRouting(deps: {active.session.state.set({...activeState,providerInstanceId: String(resolved.providerInstanceId),- ...(nextModelOptions ? { modelOptions: nextModelOptions } : {}),+ modelOptions: nextModelOptions ?? {},}),);yield* deps.runMastra("model.switch", () =>diff --git a/apps/server/src/provider/Layers/agentController/Harness.ts b/apps/server/src/provider/Layers/agentController/Harness.tsindex 1e1fd7912..9a4ac22af 100644--- a/apps/server/src/provider/Layers/agentController/Harness.ts+++ b/apps/server/src/provider/Layers/agentController/Harness.ts@@ -101,10 +101,7 @@ export function createHarness(deps: {};readonly turnStillWaiting: (threadId: string, active: ActiveSession) => boolean;readonly toolRuntime: AkeruToolRuntime;- readonly memoryUsageByThread: Map<- string,- { readonly botId: BotId; readonly capLimit: number; turnId: TurnId }- >;+ readonly memoryUsageByThread: Map<string, { readonly botId: BotId; turnId: TurnId }>;readonly readSessionStartContext: ReturnType<typeof createSessionContext>["readSessionStartContext"];@@ -353,7 +350,6 @@ export function createHarness(deps: {turnId: context.turnId,category,maximumTokens: AKERU_TURN_USAGE_RESERVATION_TOKENS,- capLimit: context.capLimit,provider: active.provider,model: active.model,createdAt: nowIso(),diff --git a/apps/server/src/provider/Layers/agentController/Policy.ts b/apps/server/src/provider/Layers/agentController/Policy.tsindex a63db8dee..69ccb35dd 100644--- a/apps/server/src/provider/Layers/agentController/Policy.ts+++ b/apps/server/src/provider/Layers/agentController/Policy.ts@@ -17,9 +17,7 @@ import {} from "@akeru/contracts";import { driverSupportsDelegation } from "@akeru/shared/delegationProviders";-import { getModelSelectionStringOptionValue } from "@akeru/shared/model";--import { getCodexServiceTierOptionValue } from "../../../codexModelOptions.ts";+import { nativeModelOptions } from "../../ReasoningOptions.ts";import { akeruToolCategory } from "../../AkeruMastraHarness.ts";@@ -48,19 +46,7 @@ export function omitNullToolFields<Input0>(input: Input0) {}export function mastraModelOptions(resolved: ResolvedEngine) {- if (resolved.provider !== "codex") return undefined;-- const reasoningEffort = getModelSelectionStringOptionValue(- resolved.modelSelection,- "reasoningEffort",- );-- const serviceTier = getCodexServiceTierOptionValue(resolved.modelSelection);-- return {- ...(reasoningEffort ? { reasoningEffort } : {}),- ...(serviceTier ? { serviceTier } : {}),- };+ return nativeModelOptions(resolved.provider, resolved.modelSelection);}export function isMissingSuspendedRun(detail: string): boolean {diff --git a/apps/server/src/provider/Layers/agentController/SessionLifecycle.ts b/apps/server/src/provider/Layers/agentController/SessionLifecycle.tsindex 1589e194b..ce698ad0c 100644--- a/apps/server/src/provider/Layers/agentController/SessionLifecycle.ts+++ b/apps/server/src/provider/Layers/agentController/SessionLifecycle.ts@@ -228,6 +228,7 @@ export function createSessionLifecycle(deps: SessionLifecycleDependencies) {existing.cwd = input.cwd;yield* deps.runMastra("state.set", () =>existing.session.state.set({+ modelOptions: deps.mastraModelOptions(resolved) ?? {},projectPath: input.cwd || undefined,yolo: false,botConversation: resolved.botConversation,@@ -582,7 +583,7 @@ export function createSessionLifecycle(deps: SessionLifecycleDependencies) {...(input.botName ? { botName: input.botName } : {}),personalityTone,mcpInstructions: formatMcpServerInstructions(mcpServers),- ...(modelOptions ? { modelOptions } : {}),+ modelOptions: modelOptions ?? {},}),);yield* deps.runMastra("model.switch", () =>diff --git a/apps/server/src/provider/Layers/agentController/SessionLifecycleDependencies.ts b/apps/server/src/provider/Layers/agentController/SessionLifecycleDependencies.tsindex fd6e80a76..9cf8b485e 100644--- a/apps/server/src/provider/Layers/agentController/SessionLifecycleDependencies.ts+++ b/apps/server/src/provider/Layers/agentController/SessionLifecycleDependencies.ts@@ -1,3 +1,4 @@+import type { AkeruModelOptions } from "../../ReasoningOptions.ts";import type { AkeruControllerHarness } from "../../mastra/AkeruHarnessTypes.ts";import { ProviderDriverKind } from "@akeru/contracts";import { ProviderInstanceId } from "@akeru/contracts";@@ -150,9 +151,7 @@ export interface SessionLifecycleDependencies {clearedDisabledFor: { id: BotId; name: string }[];}>;readonly bundle: AkeruControllerHarness;- readonly mastraModelOptions: (- resolved: ResolvedEngine,- ) => { serviceTier?: string; reasoningEffort?: string } | undefined;+ readonly mastraModelOptions: (resolved: ResolvedEngine) => AkeruModelOptions | undefined;readonly DEFAULT_MODE_ID: string;readonly permissionPolicy: (runtimeMode: RuntimeMode,diff --git a/apps/server/src/provider/Layers/agentController/TurnLifecycle.ts b/apps/server/src/provider/Layers/agentController/TurnLifecycle.tsindex 5803dabcf..7d0803d0f 100644--- a/apps/server/src/provider/Layers/agentController/TurnLifecycle.ts+++ b/apps/server/src/provider/Layers/agentController/TurnLifecycle.ts@@ -74,10 +74,7 @@ export function createTurnLifecycle(deps: {createdAt: string;};readonly toolRuntime: AkeruToolRuntime;- readonly memoryUsageByThread: Map<- string,- { readonly botId: BotId; readonly capLimit: number; turnId: TurnId }- >;+ readonly memoryUsageByThread: Map<string, { readonly botId: BotId; turnId: TurnId }>;readonly publishSessionState: (threadId: ThreadId,active: ActiveSession,diff --git a/apps/server/src/provider/Layers/opencode/OpenCodeProtocol.ts b/apps/server/src/provider/Layers/opencode/OpenCodeProtocol.tsindex c7e42d349..4af43837f 100644--- a/apps/server/src/provider/Layers/opencode/OpenCodeProtocol.ts+++ b/apps/server/src/provider/Layers/opencode/OpenCodeProtocol.ts@@ -32,6 +32,12 @@ import {type OpenCodeSessionContext,} from "./OpenCodeAdapterState.ts";+/**+ * The variant choice that leaves OpenCode on its own configured default. It is+ * namespaced so it never matches a variant name a user configured.+ */+export const OPENCODE_PROVIDER_DEFAULT_VARIANT = "akeru:provider-default";+/*** Decode a persisted resume cursor into the upstream `ses_…` id. Anything* that isn't a current-version cursor with a non-empty id means "no resume"diff --git a/apps/server/src/provider/Layers/test-support/projectionFixtures.ts b/apps/server/src/provider/Layers/test-support/projectionFixtures.tsindex ba2a36a1e..1db1d29b0 100644--- a/apps/server/src/provider/Layers/test-support/projectionFixtures.ts+++ b/apps/server/src/provider/Layers/test-support/projectionFixtures.ts@@ -29,7 +29,6 @@ export function providerBotFixture(engine: null,sandbox: "local",runtimeMode: "full-access",- usageCap: null,imageProvider: null,voiceEnabled: false,channelBindings: [],diff --git a/apps/server/src/provider/ModelCatalog.ts b/apps/server/src/provider/ModelCatalog.tsindex 1e1c64847..b33f6c26f 100644--- a/apps/server/src/provider/ModelCatalog.ts+++ b/apps/server/src/provider/ModelCatalog.ts@@ -1,3 +1,4 @@+import { reasoningCapabilities } from "./ReasoningOptions.ts";/*** ModelCatalog — the provider models Akeru offers, their display names, and* which of them are current versus legacy.@@ -136,7 +137,13 @@ export function catalogProviderModels(input: {name: names.get(slug) ?? slug,isCustom,...(slug === defaultSlug ? { isDefault: true } : {}),- capabilities: null,+ capabilities: isCustom+ ? null+ : reasoningCapabilities(+ input.driver,+ slug,+ catalogModelsFor(input.catalog, input.driver).find((entry) => entry.id === slug),+ ),})),input.catalog,input.driver,diff --git a/apps/server/src/provider/ReasoningOptions.test.ts b/apps/server/src/provider/ReasoningOptions.test.tsnew file mode 100644index 000000000..d6c7f8dae--- /dev/null+++ b/apps/server/src/provider/ReasoningOptions.test.ts@@ -0,0 +1,425 @@+import * as Predicate from "effect/Predicate";+import { describe, expect, it, vi } from "vite-plus/test";+import { ProviderDriverKind, ProviderInstanceId, type ModelSelection } from "@akeru/contracts";+import { AuthStorage } from "@mastra/code-sdk/auth/storage";+import { RequestContext } from "@mastra/core/request-context";+import {+ nativeModelOptions,+ reasoningCapabilities,+ invalidReasoningSelection,+} from "./ReasoningOptions.ts";+import { claudeHarnessCapabilities } from "./ClaudeReasoningCapabilities.ts";+import {+ mastraModelId,+ resolveAkeruMastraModel,+ withAkeruModelRunOptions,+} from "./mastra/AkeruModels.ts";+import { controllerModelOptions } from "./mastra/AkeruMemory.ts";++const selection = (model: string, options?: ModelSelection["options"]): ModelSelection => ({+ instanceId: ProviderInstanceId.make("test"),+ model,+ ...(options ? { options } : {}),+});++const driver = ProviderDriverKind.make;++const auth = new AuthStorage("/tmp/akeru-reasoning-unused-auth.json");++const stopped = new Error("request captured");++async function capturedRequest(+ modelId: string,+ provider: ProviderDriverKind,+ options?: ModelSelection["options"],+ oauth = false,+) {+ let body: unknown;++ const fetch = vi.spyOn(globalThis, "fetch").mockImplementation(async (_input, init) => {+ body = JSON.parse(String(init?.body));+ throw stopped;+ });++ vi.spyOn(auth, "reload").mockImplementation(() => {});+ vi.spyOn(auth, "get").mockReturnValue({+ type: "oauth",+ access: "test-token",+ refresh: "refresh",+ expires: Number.MAX_SAFE_INTEGER,+ });+ vi.spyOn(auth, "getApiKey").mockResolvedValue("test-token");++ try {+ const modelOptions = nativeModelOptions(provider, selection(modelId, options));++ const model = resolveAkeruMastraModel(+ mastraModelId(provider, modelId),+ auth,+ async () => ({ accessToken: "kimi-token" }),+ async () => "go-key",+ modelOptions,+ oauth ? undefined : () => ({ type: "api-key", access: "test-key" }),+ );++ if (+ !Predicate.isObject(model) ||+ !("specificationVersion" in model) ||+ model.specificationVersion !== "v3"+ )+ throw new Error("Expected a native language model");++ const run = withAkeruModelRunOptions({}, modelOptions ? { modelOptions } : {});++ await expect(+ model.doGenerate({+ prompt: [{ role: "user", content: [{ type: "text", text: "hello" }] }],+ ...run,+ }),+ ).rejects.toThrow(stopped);+ expect(fetch).toHaveBeenCalledOnce();++ return body;+ } finally {+ vi.restoreAllMocks();+ }+}++describe("native reasoning requests", () => {+ it.each([false, true])(+ "Codex API/OAuth preserves minimal and leaves absent effort to the provider (%s)",+ async (oauth) => {+ expect(+ await capturedRequest(+ "gpt-5",+ driver("codex"),+ [{ id: "reasoningEffort", value: "minimal" }],+ oauth,+ ),+ ).toMatchObject({ reasoning: { effort: "minimal" } });++ const defaultRequest = await capturedRequest(+ "gpt-5",+ driver("codex"),+ [{ id: "reasoningEffort", value: "default" }],+ oauth,+ );++ if (oauth)+ expect(defaultRequest).toMatchObject({+ reasoning: { effort: "medium" },+ store: false,+ instructions: expect.any(String),+ });+ else expect(defaultRequest).not.toHaveProperty("reasoning");+ expect(+ await capturedRequest(+ "gpt-5",+ driver("codex"),+ [{ id: "reasoningEffort", value: "off" }],+ oauth,+ ),+ ).toMatchObject({ reasoning: { effort: "none" } });+ },+ );+ it.each([false, true])(+ "Claude API/OAuth sends native effort and explicit thinking false (%s)",+ async (oauth) => {+ expect(+ await capturedRequest(+ "claude-opus-5-5",+ driver("claudeAgent"),+ [{ id: "effort", value: "xhigh" }],+ oauth,+ ),+ ).toMatchObject({ output_config: { effort: "xhigh" }, thinking: { type: "adaptive" } });+ expect(+ await capturedRequest(+ "claude-haiku-4-5",+ driver("claudeAgent"),+ [{ id: "thinking", value: false }],+ oauth,+ ),+ ).toMatchObject({ thinking: { type: "disabled" } });+ expect(+ await capturedRequest("claude-opus-5-5", driver("claudeAgent"), undefined, oauth),+ ).not.toHaveProperty("thinking");+ },+ );+ it.each([false, true])(+ "Grok API/OAuth sends the compatible native effort (%s)",+ async (oauth) => {+ expect(+ await capturedRequest(+ "grok-4.7",+ driver("grok"),+ [{ id: "reasoningEffort", value: "high" }],+ oauth,+ ),+ ).toMatchObject({ reasoning_effort: "high" });+ },+ );+ it.each([false, true])(+ "Haiku API/OAuth toggle uses manual thinking with the SDK minimum (%s)",+ async (oauth) => {+ expect(+ await capturedRequest(+ "claude-haiku-4-5",+ driver("claudeAgent"),+ [{ id: "thinking", value: true }],+ oauth,+ ),+ ).toMatchObject({ thinking: { type: "enabled", budget_tokens: 1024 } });+ },+ );+ it.each([false, true])(+ "Opus 4.5 API/OAuth sends effort without adaptive thinking (%s)",+ async (oauth) => {+ const body = await capturedRequest(+ "claude-opus-4-5",+ driver("claudeAgent"),+ [{ id: "effort", value: "low" }],+ oauth,+ );++ expect(body).toMatchObject({ output_config: { effort: "low" } });+ expect(body).not.toHaveProperty("thinking");+ },+ );+ it("Custom API sends the saved effort under its own provider options", async () => {+ let body: unknown;++ const fetch = vi.spyOn(globalThis, "fetch").mockImplementation(async (_input, init) => {+ body = JSON.parse(String(init?.body));+ throw stopped;+ });++ try {+ const modelOptions = nativeModelOptions(+ driver("customOpenai"),+ selection("local-model", [{ id: "reasoningEffort", value: "high" }]),+ );++ expect(modelOptions?.namespace).toBe("custom-openai");++ const model = resolveAkeruMastraModel(+ "custom-openai/local-model",+ auth,+ undefined,+ undefined,+ modelOptions,+ undefined,+ {+ environment: {},+ instanceEnvironment: { CUSTOM_OPENAI_BASE_URL: "http://127.0.0.1:9/v1" },+ useSavedCredential: false,+ },+ );++ if (+ !Predicate.isObject(model) ||+ !("specificationVersion" in model) ||+ model.specificationVersion !== "v3"+ )+ throw new Error("Expected a native language model");++ await expect(+ model.doGenerate({+ prompt: [{ role: "user", content: [{ type: "text", text: "hello" }] }],+ ...withAkeruModelRunOptions({}, modelOptions ? { modelOptions } : {}),+ }),+ ).rejects.toThrow(stopped);+ expect(fetch).toHaveBeenCalledOnce();+ expect(body).toMatchObject({ reasoning_effort: "high" });+ } finally {+ vi.restoreAllMocks();+ }+ });++ it("Kimi sends adaptive effort without inventing a budget", async () => {+ const body = await capturedRequest("kimi-for-coding", driver("kimi"), [+ { id: "effort", value: "max" },+ ]);++ expect(body).toMatchObject({+ output_config: { effort: "max" },+ thinking: { type: "adaptive" },+ });+ expect(body).not.toHaveProperty("thinking.budget_tokens");+ });+ it.each([+ [+ "muse-spark-1.3-contributor",+ "reasoningEffort",+ "minimal",+ { reasoning: { effort: "minimal" } },+ ],+ ["qwen3.8-max", "effort", "xhigh", { output_config: { effort: "xhigh" } }],+ ["deepseek-v4-pro", "reasoningEffort", "max", { reasoning_effort: "max" }],+ ] as const)("Go %s serializes its protocol's field", async (model, id, value, expected) => {+ expect(await capturedRequest(model, driver("opencodeGo"), [{ id, value }])).toMatchObject(+ expected,+ );+ });+});++describe("reasoning capability and state boundaries", () => {+ it("retains only catalog levels and gives fixed models no invented levels", () => {+ const entry = { id: "model", name: "Model", efforts: ["minimal", "high"] };+ expect(+ reasoningCapabilities(driver("opencodeGo"), "muse-spark-1.3-contributor", entry)+ .optionDescriptors?.[0],+ ).toMatchObject({ options: [{ id: "default" }, { id: "minimal" }, { id: "high" }] });+ expect(+ reasoningCapabilities(driver("kimi"), "kimi-for-coding-highspeed", undefined)+ .optionDescriptors,+ ).toEqual([]);+ });+ it("separates native Claude effort from workflow choices", () => {+ const caps = claudeHarnessCapabilities("claude-opus-5-5", {+ id: "claude-opus-5-5",+ name: "Opus",+ efforts: ["low", "medium", "high", "xhigh", "max"],+ });++ expect(caps.optionDescriptors?.find((option) => option.id === "effort")).toMatchObject({+ options: [+ { id: "default" },+ { id: "low" },+ { id: "medium" },+ { id: "high" },+ { id: "xhigh" },+ { id: "max" },+ ],+ });+ expect(caps.optionDescriptors?.find((option) => option.id === "workflow")).toBeUndefined();+ expect(caps.optionDescriptors?.find((option) => option.id === "thinking")).toBeUndefined();+ expect(claudeHarnessCapabilities("claude-haiku-4-5", undefined).optionDescriptors).toEqual(+ expect.arrayContaining([expect.objectContaining({ id: "thinking" })]),+ );+ });++ it("retains known historical Claude effort when the refreshed catalog omits the model", () => {+ const caps = claudeHarnessCapabilities("claude-opus-4-5", undefined);++ expect(caps.optionDescriptors?.find((option) => option.id === "effort")).toMatchObject({+ options: [{ id: "default" }, { id: "low" }, { id: "medium" }, { id: "high" }],+ });+ expect(+ invalidReasoningSelection(+ selection("claude-opus-4-5", [{ id: "effort", value: "low" }]),+ caps,+ ),+ ).toBeUndefined();+ expect(claudeHarnessCapabilities("unknown-model", undefined).optionDescriptors).toEqual([]);+ expect(+ claudeHarnessCapabilities("claude-opus-4-5", {+ id: "claude-opus-4-5",+ name: "Opus",+ }).optionDescriptors?.find((option) => option.id === "effort"),+ ).toBeUndefined();+ });+ it("keeps the legacy Codex Fast toggle as the fast service tier", () => {+ expect(+ nativeModelOptions(driver("codex"), selection("gpt-5", [{ id: "fastMode", value: true }])),+ ).toEqual({ serviceTier: "fast" });+ expect(+ nativeModelOptions(+ driver("codex"),+ selection("gpt-5", [+ { id: "serviceTier", value: "priority" },+ { id: "fastMode", value: true },+ ]),+ ),+ ).toEqual({ serviceTier: "priority" });+ });++ it("runs saved Claude workflow efforts at provider default instead of rejecting them", () => {+ const caps = claudeHarnessCapabilities("claude-opus-5-5", undefined);++ for (const value of ["ultrathink", "ultracode"]) {+ const saved = selection("claude-opus-5-5", [{ id: "effort", value }]);++ expect(invalidReasoningSelection(saved, caps)).toBeUndefined();+ expect(nativeModelOptions(driver("claudeAgent"), saved)?.effort).toBeUndefined();+ }+ });+ it("decodes stored native fields including explicit false, and old Codex fields", () => {+ const context = new RequestContext();++ for (const modelOptions of [+ { namespace: "anthropic", effort: "max", thinking: false },+ { reasoningEffort: "high", serviceTier: "priority" },+ ] as const) {+ context.setRaw("controller", { state: { modelOptions } });+ expect(controllerModelOptions(context)).toEqual(modelOptions);+ }+ });+ it("validates known reasoning descriptors without rejecting unrelated or unsettled selections", () => {+ const caps = reasoningCapabilities(driver("kimi"), "kimi-for-coding", {+ id: "kimi-for-coding",+ name: "Kimi",+ efforts: ["low", "max"],+ });++ expect(+ invalidReasoningSelection(+ selection("kimi-for-coding", [{ id: "effort", value: "medium" }]),+ caps,+ ),+ ).toContain("does not support");+ expect(+ invalidReasoningSelection(+ selection("kimi-for-coding", [{ id: "custom", value: "future" }]),+ caps,+ ),+ ).toBeUndefined();+ expect(+ invalidReasoningSelection(+ selection("kimi-for-coding", [{ id: "effort", value: "medium" }]),+ null,+ ),+ ).toBeUndefined();+ const fixed = reasoningCapabilities(driver("kimi"), "fixed", undefined);+ expect(+ invalidReasoningSelection(selection("fixed", [{ id: "effort", value: "low" }]), fixed),+ ).toContain("not supported");+ expect(+ invalidReasoningSelection(selection("fixed", [{ id: "effort", value: "default" }]), fixed),+ ).toBeUndefined();++ const codexCaps = reasoningCapabilities(driver("codex"), "gpt-5", {+ id: "gpt-5",+ name: "GPT 5",+ efforts: ["minimal", "low", "medium", "high"],+ });++ expect(+ invalidReasoningSelection(+ selection("gpt-5", [{ id: "reasoningEffort", value: "off" }]),+ codexCaps,+ ),+ ).toContain("does not support");+ expect(+ invalidReasoningSelection(selection("gpt-5", [{ id: "reasoningEffort", value: "off" }]), {+ optionDescriptors: [+ {+ id: "reasoningEffort",+ label: "Reasoning",+ type: "select",+ options: [{ id: "off", label: "Off" }],+ },+ ],+ }),+ ).toBeUndefined();+ expect(nativeModelOptions(driver("kimi"), selection("kimi-for-coding"))).toBeUndefined();+ expect(+ withAkeruModelRunOptions({}, { modelOptions: { effort: "high", thinking: false } }),+ ).toEqual({+ providerOptions: { anthropic: { effort: "high", thinking: { type: "disabled" } } },+ });+ expect(withAkeruModelRunOptions({}, { modelOptions: { reasoningEffort: "default" } })).toEqual({+ providerOptions: { openai: {} },+ });+ });+});diff --git a/apps/server/src/provider/ReasoningOptions.ts b/apps/server/src/provider/ReasoningOptions.tsnew file mode 100644index 000000000..2697e35a7--- /dev/null+++ b/apps/server/src/provider/ReasoningOptions.ts@@ -0,0 +1,184 @@+import * as Predicate from "effect/Predicate";+import {+ type ModelCapabilities,+ type ModelSelection,+ type ProviderDriverKind,+} from "@akeru/contracts";+import {+ createModelCapabilities,+ getModelSelectionStringOptionValue,+ getModelSelectionBooleanOptionValue,+} from "@akeru/shared/model";+import * as Schema from "effect/Schema";+import { getCodexServiceTierOptionValue } from "../codexModelOptions.ts";+import { openCodeGoProtocol } from "./AkeruOpenCodeGoProvider.ts";+import { getClaudeModelCapabilities } from "./Layers/claude/ClaudeModels.ts";+import { type CatalogModel } from "./modelCatalogData.ts";++export const AkeruModelOptions = Schema.Struct({+ namespace: Schema.optionalKey(+ Schema.Literals(["openai", "anthropic", "xai", "opencode-go", "custom-openai"]),+ ),+ reasoningEffort: Schema.optionalKey(Schema.String),+ serviceTier: Schema.optionalKey(Schema.String),+ effort: Schema.optionalKey(Schema.String),+ thinking: Schema.optionalKey(Schema.Boolean),+ thinkingMode: Schema.optionalKey(Schema.Literals(["adaptive", "enabled"])),+});++export type AkeruModelOptions = typeof AkeruModelOptions.Type;++const NATIVE_EFFORTS = new Set(["none", "minimal", "low", "medium", "high", "xhigh", "max"]);++/** Claude CLI workflow modes older bots may have saved; harness turns run them at provider default. */+const LEGACY_CLAUDE_WORKFLOW_EFFORTS = new Set(["ultrathink", "ultracode"]);++export function reasoningCapabilities(+ driver: ProviderDriverKind,+ slug: string,+ entry: CatalogModel | undefined,+): ModelCapabilities {+ const levels = entry?.efforts?.filter((level) => NATIVE_EFFORTS.has(level)) ?? [];++ const anthropic =+ driver === "claudeAgent" ||+ driver === "kimi" ||+ (driver === "opencodeGo" && openCodeGoProtocol(slug) === "anthropic");++ return createModelCapabilities({+ optionDescriptors:+ levels.length === 0+ ? []+ : [+ {+ id: anthropic ? "effort" : "reasoningEffort",+ label: "Reasoning",+ type: "select",+ currentValue: "default",+ options: [+ { id: "default", label: "Provider default", isDefault: true },+ ...levels.map((id) => ({+ id,+ label: id === "xhigh" ? "Extra High" : id[0]!.toUpperCase() + id.slice(1),+ })),+ ],+ },+ ],+ });+}++export function nativeModelOptions(+ driver: ProviderDriverKind,+ selection: ModelSelection,+): AkeruModelOptions | undefined {+ const reasoningEffort = getModelSelectionStringOptionValue(selection, "reasoningEffort");+ const effort = getModelSelectionStringOptionValue(selection, "effort");++ const thinking =+ driver === "claudeAgent" &&+ getClaudeModelCapabilities(selection.model).optionDescriptors?.some(+ (descriptor) => descriptor.id === "thinking",+ )+ ? getModelSelectionBooleanOptionValue(selection, "thinking")+ : undefined;++ const serviceTier = driver === "codex" ? getCodexServiceTierOptionValue(selection) : undefined;++ const anthropic =+ driver === "claudeAgent" ||+ driver === "kimi" ||+ (driver === "opencodeGo" && openCodeGoProtocol(selection.model) === "anthropic");++ const namespace = anthropic+ ? "anthropic"+ : driver === "grok"+ ? "xai"+ : driver === "customOpenai"+ ? "custom-openai"+ : driver === "opencodeGo" && openCodeGoProtocol(selection.model) === "chat-completions"+ ? "opencode-go"+ : "openai";++ const rawEffort = anthropic ? (effort ?? reasoningEffort) : reasoningEffort;+ const normalized = rawEffort === "off" && !anthropic ? "none" : rawEffort;+ const nativeEffort = normalized && NATIVE_EFFORTS.has(normalized) ? normalized : undefined;++ const adaptive =+ driver === "kimi" || (driver === "claudeAgent" && supportsAdaptiveThinking(selection.model));++ if (!nativeEffort && thinking === undefined && !serviceTier) return undefined;++ return {+ ...(driver === "codex" ? {} : { namespace }),+ ...(nativeEffort+ ? anthropic+ ? { effort: nativeEffort }+ : { reasoningEffort: nativeEffort }+ : {}),+ ...(thinking !== undefined+ ? { thinking, thinkingMode: "enabled" as const }+ : nativeEffort && adaptive+ ? { thinking: true, thinkingMode: "adaptive" as const }+ : {}),+ ...(serviceTier ? { serviceTier } : {}),+ };+}++export function supportsAdaptiveThinking(slug: string): boolean {+ const version = /^claude-(?:opus|sonnet|fable)-(\d+)(?:-(\d+))?/.exec(slug);++ return (+ version !== null &&+ (Number(version[1]) > 4 || (Number(version[1]) === 4 && Number(version[2]) >= 6))+ );+}++/** Older snapshots without descriptors cannot authoritatively reject a saved selection. */+export function invalidReasoningSelection(+ selection: ModelSelection,+ capabilities: ModelCapabilities | null | undefined,+): string | undefined {+ for (const option of selection.options ?? []) {+ if (!["reasoningEffort", "effort", "thinking"].includes(option.id)) continue;++ const descriptor = capabilities?.optionDescriptors?.find(+ (candidate) => candidate.id === option.id,+ );++ if (option.value === "default" && option.id !== "thinking") continue;++ if (+ option.id === "effort" &&+ Predicate.isString(option.value) &&+ LEGACY_CLAUDE_WORKFLOW_EFFORTS.has(option.value)+ ) {+ continue;+ }++ if (!descriptor) {+ if (capabilities?.optionDescriptors === undefined) continue;++ return `Reasoning option '${option.id}' is not supported for '${selection.model}'.`;+ }++ // `off` is the stored Codex spelling of the native `none` level.+ const value =+ option.id === "reasoningEffort" &&+ option.value === "off" &&+ descriptor.type === "select" &&+ descriptor.options.some((candidate) => candidate.id === "none")+ ? "none"+ : option.value;++ if (+ descriptor.type === "boolean"+ ? !Predicate.isBoolean(value)+ : !Predicate.isString(value) ||+ !descriptor.options.some((candidate) => candidate.id === value)+ ) {+ return `Reasoning option '${option.id}' does not support '${String(option.value)}' for '${selection.model}'.`;+ }+ }++ return undefined;+}diff --git a/apps/server/src/provider/Services/AgentController.ts b/apps/server/src/provider/Services/AgentController.tsindex 31b117e0a..3a86ffa84 100644--- a/apps/server/src/provider/Services/AgentController.ts+++ b/apps/server/src/provider/Services/AgentController.ts@@ -45,7 +45,6 @@ export interface AgentControllerEngineSelection extends AgentControllerAvailableexport type AgentControllerSendTurnInput = ProviderSendTurnInput & {readonly botUsage?: {readonly botId: BotId;- readonly capLimit: number;};/*** Finished child work acknowledged for this turn, formatted as context. Eachdiff --git a/apps/server/src/provider/delegation/AkeruDelegationControls.ts b/apps/server/src/provider/delegation/AkeruDelegationControls.tsindex 9bcb313ec..7f984fcdd 100644--- a/apps/server/src/provider/delegation/AkeruDelegationControls.ts+++ b/apps/server/src/provider/delegation/AkeruDelegationControls.ts@@ -78,7 +78,6 @@ export function createAkeruDelegationControls(engine: parentBot.engine,sandbox: parentBot.sandbox,runtimeMode: parent.access.runtimeMode,- usageCap: null,imageProvider: null,voiceEnabled: false,groupId: parentThread.groupId ?? null,diff --git a/apps/server/src/provider/mastra/AkeruHarnessTypes.ts b/apps/server/src/provider/mastra/AkeruHarnessTypes.tsindex 1056695f1..a8dcebfba 100644--- a/apps/server/src/provider/mastra/AkeruHarnessTypes.ts+++ b/apps/server/src/provider/mastra/AkeruHarnessTypes.ts@@ -19,6 +19,8 @@ import {type AkeruRoutineDeleteResult,} from "./AkeruRoutineSchemas.ts";+import type { AkeruModelOptions } from "../ReasoningOptions.ts";+export interface AkeruMastraState {readonly providerInstanceId?: string;// Undefined clears a previous directory when a reused session loses its cwd.@@ -29,10 +31,7 @@ export interface AkeruMastraState {readonly personalityTone?: BotPersonalityTone;readonly persistentMemoryContext?: string;readonly mcpInstructions?: string;- readonly modelOptions?: {- readonly reasoningEffort?: string;- readonly serviceTier?: string;- };+ readonly modelOptions?: AkeruModelOptions;}export type AkeruMastraSession = Session<AkeruMastraState>;diff --git a/apps/server/src/provider/mastra/AkeruMemory.ts b/apps/server/src/provider/mastra/AkeruMemory.tsindex d63854885..e74b05499 100644--- a/apps/server/src/provider/mastra/AkeruMemory.ts+++ b/apps/server/src/provider/mastra/AkeruMemory.ts@@ -1,3 +1,4 @@+import { AkeruModelOptions } from "../ReasoningOptions.ts";import * as Schema from "effect/Schema";import * as Option from "effect/Option";import * as Predicate from "effect/Predicate";@@ -65,12 +66,7 @@ const decodeControllerContext = Schema.decodeUnknownOption(}),);-const decodeModelOptions = Schema.decodeUnknownOption(- Schema.Struct({- reasoningEffort: Schema.optionalKey(Schema.String),- serviceTier: Schema.optionalKey(Schema.String),- }),-);+const decodeModelOptions = Schema.decodeUnknownOption(AkeruModelOptions);export function controllerContext(requestContext: RequestContext) {return Option.getOrUndefined(decodeControllerContext(requestContext.getRaw("controller")));diff --git a/apps/server/src/provider/mastra/AkeruModels.ts b/apps/server/src/provider/mastra/AkeruModels.tsindex 9df9d7389..4eafbd2c2 100644--- a/apps/server/src/provider/mastra/AkeruModels.ts+++ b/apps/server/src/provider/mastra/AkeruModels.ts@@ -2,14 +2,12 @@ import * as Schema from "effect/Schema";import * as Predicate from "effect/Predicate";import { AuthStorage } from "@mastra/code-sdk/auth/storage";import { opencodeClaudeMaxProvider } from "@mastra/code-sdk/providers/claude-max";-import { openaiCodexProvider } from "@mastra/code-sdk/providers/openai-codex";import { xaiProvider } from "@mastra/code-sdk/providers/xai";import { createAnthropic } from "@ai-sdk/anthropic";import { createOpenAICompatible } from "@ai-sdk/openai-compatible";-import { isThinkingLevelSetting } from "@mastra/code-sdk/thinking";import { SubscriptionProviderId, type ProviderDriverKind } from "@akeru/contracts";import type { SubscriptionAuthService } from "../../subscription-auth/service.ts";-import { akeruOpenAIProvider } from "../AkeruOpenAIProvider.ts";+import { akeruCodexOAuthProvider, akeruOpenAIProvider } from "../AkeruOpenAIProvider.ts";import { akeruKimiProvider, type AkeruKimiAccess } from "../AkeruKimiProvider.ts";import { akeruOpenCodeGoProvider } from "../AkeruOpenCodeGoProvider.ts";import { type AkeruMastraState } from "./AkeruHarnessTypes.ts";@@ -31,29 +29,46 @@ export function withAkeruModelRunOptions<Options extends AkeruRunOptions>(runOptions: Options,state: AkeruMastraState,) {- const serviceTier = state.modelOptions?.serviceTier;+ const modelOptions = state.modelOptions;- if (!serviceTier) return runOptions;+ if (!modelOptions) return runOptions;- const providerOptions =- Predicate.isObjectOrArray(runOptions.providerOptions) && runOptions.providerOptions !== null- ? runOptions.providerOptions- : {};+ const namespace =+ modelOptions.namespace ??+ (modelOptions.effort !== undefined || modelOptions.thinking !== undefined+ ? "anthropic"+ : "openai");- const openai =- "openai" in providerOptions &&- Predicate.isObjectOrArray(providerOptions.openai) &&- providerOptions.openai !== null- ? providerOptions.openai- : {};+ const reasoningEffort =+ modelOptions.reasoningEffort === "off" ? "none" : modelOptions.reasoningEffort;++ const providerOptions = Predicate.isObject(runOptions.providerOptions)+ ? runOptions.providerOptions+ : {};++ const existing = providerOptions[namespace];++ const priorOptions = Predicate.isObject(existing) ? existing : {};- return {- ...runOptions,- providerOptions: {- ...providerOptions,- openai: { ...openai, serviceTier },- },+ const native = {+ ...priorOptions,+ ...(reasoningEffort && reasoningEffort !== "default"+ ? { reasoningEffort, ...(namespace === "openai" ? { forceReasoning: true } : {}) }+ : {}),+ ...(modelOptions.serviceTier ? { serviceTier: modelOptions.serviceTier } : {}),+ ...(modelOptions.effort && modelOptions.effort !== "default"+ ? { effort: modelOptions.effort }+ : {}),+ ...(modelOptions.thinking !== undefined+ ? {+ thinking: {+ type: modelOptions.thinking ? (modelOptions.thinkingMode ?? "adaptive") : "disabled",+ },+ }+ : {}),};++ return { ...runOptions, providerOptions: { ...providerOptions, [namespace]: native } };}export const MASTRA_MODEL_PREFIX = {@@ -112,7 +127,7 @@ export function resolveAkeruMastraModel(authStorage: AuthStorage,getKimiAccess?: (instanceId?: string) => Promise<AkeruKimiAccess | undefined>,getOpenCodeGoApiKey?: (instanceId?: string) => Promise<string | undefined>,- modelOptions?: AkeruMastraState["modelOptions"],+ _modelOptions?: AkeruMastraState["modelOptions"],getSubscriptionApiKey?: SubscriptionAuthService["getApiKeyCredential"],connection?: {readonly environment: NodeJS.ProcessEnv;@@ -174,12 +189,7 @@ export function resolveAkeruMastraModel(throw new Error("This Codex instance has no OPENAI_API_KEY transport for Akeru Mastra.");}- const reasoningEffort = modelOptions?.reasoningEffort;-- return openaiCodexProvider(trimmed.slice("openai/".length), {- authStorage: scopedAuthStorage,- ...(isThinkingLevelSetting(reasoningEffort) ? { thinkingLevel: reasoningEffort } : {}),- });+ return akeruCodexOAuthProvider(trimmed.slice("openai/".length), scopedAuthStorage);}if (trimmed.startsWith("anthropic/")) {diff --git a/apps/server/src/provider/testUtils/delegationHarness.ts b/apps/server/src/provider/testUtils/delegationHarness.tsindex 4e708f16e..3067d5363 100644--- a/apps/server/src/provider/testUtils/delegationHarness.ts+++ b/apps/server/src/provider/testUtils/delegationHarness.ts@@ -67,7 +67,6 @@ export function bot(id: BotId, overrides: Partial<OrchestrationBot> = {}): Orcheengine: null,sandbox: "local",runtimeMode: "approval-required",- usageCap: null,imageProvider: null,voiceEnabled: false,channelBindings: [],diff --git a/apps/server/src/routines/RoutineDraftDispatcher.test.ts b/apps/server/src/routines/RoutineDraftDispatcher.test.tsindex 4248653d7..f2d9a65d1 100644--- a/apps/server/src/routines/RoutineDraftDispatcher.test.ts+++ b/apps/server/src/routines/RoutineDraftDispatcher.test.ts@@ -50,7 +50,6 @@ function bot(id: BotId): OrchestrationBot {engine: null,sandbox: "local",runtimeMode: "full-access",- usageCap: null,imageProvider: null,voiceEnabled: false,channelBindings: [],diff --git a/apps/server/src/routines/RuntimeAdapterLive.test.ts b/apps/server/src/routines/RuntimeAdapterLive.test.tsindex c104f086b..987cbc185 100644--- a/apps/server/src/routines/RuntimeAdapterLive.test.ts+++ b/apps/server/src/routines/RuntimeAdapterLive.test.ts@@ -90,7 +90,6 @@ const bot = (botId: string, name: string, provider: string): ProjectionBot => ({avatar: { kind: "blob", shape: "circle", color: "blue" },sandbox: "local",runtimeMode: "approval-required",- usageCap: null,imageProvider: null,voiceEnabled: false,groupId: null,diff --git a/apps/server/src/routines/RuntimeLive.restart.test.ts b/apps/server/src/routines/RuntimeLive.restart.test.tsindex e82f7001e..32beae7a1 100644--- a/apps/server/src/routines/RuntimeLive.restart.test.ts+++ b/apps/server/src/routines/RuntimeLive.restart.test.ts@@ -160,7 +160,6 @@ const seed = Effect.gen(function* () {engine: null,sandbox: "local",runtimeMode: "approval-required",- usageCap: null,groupId: null,createdAt: CREATED_AT,});diff --git a/apps/server/src/server.admission.test.ts b/apps/server/src/server.admission.test.tsindex 442a78d43..7248fce06 100644--- a/apps/server/src/server.admission.test.ts+++ b/apps/server/src/server.admission.test.ts@@ -3,7 +3,6 @@ import * as Predicate from "effect/Predicate";import * as NodeHttpServer from "@effect/platform-node/NodeHttpServer";import * as NodeServices from "@effect/platform-node/NodeServices";import {- BotId,CommandId,DEFAULT_SERVER_SETTINGS,MessageId,@@ -21,7 +20,6 @@ import { HttpBody, HttpClient } from "effect/unstable/http";import { vi } from "vite-plus/test";import * as ServerConfig from "./config.ts";import * as OrchestrationEngine from "./orchestration/Services/OrchestrationEngine.ts";-import * as ProjectionBots from "./persistence/Services/ProjectionBots.ts";import { buildAppUnderTest } from "./serverTestApp.ts";import {@@ -29,7 +27,6 @@ import {defaultProjectId,defaultModelSelection,defaultDesktopBootstrapToken,- makeChannelTestBot,} from "./serverTestFixtures.ts";import { getWsServerUrl, withWsRpcClient, exchangeAccessToken } from "./serverTestClients.ts";@@ -312,175 +309,4 @@ it.layer(NodeServices.layer)("server router seam", (it) => {assert.equal(dispatch.mock.calls.length, 1);}).pipe(Effect.provide(NodeHttpServer.layerTest)),);-- it.effect("rejects a capped bot chat through HTTP before dispatch", () =>- Effect.gen(function* () {- const botId = BotId.make("bot-http-bootstrap-capped");-- const bot = {- ...makeChannelTestBot(),- botId,- imageProvider: null,- usageCap: { unit: "tokens" as const, limit: 100 },- } satisfies ProjectionBots.ProjectionBot;-- const dispatch = vi.fn<OrchestrationEngine.OrchestrationEngineService["Service"]["dispatch"]>(- () => Effect.succeed({ sequence: 1 }),- );-- yield* buildAppUnderTest({- layers: {- providerRegistry: { getProviders: Effect.succeed([readyDefaultProvider]) },- projectionBots: {- getById: ({ botId: requested }) =>- Effect.succeed(requested === botId ? Option.some(bot) : Option.none()),- },- botUsageLedger: {- summarize: () =>- Effect.succeed({- botId,- consumedTokens: 100,- reservedTokens: 0,- measurements: {- input: { tokens: 100, unavailableEntries: 0 },- output: { tokens: 0, unavailableEntries: 0 },- observer: { tokens: 0, unavailableEntries: 0 },- reflector: { tokens: 0, unavailableEntries: 0 },- },- entries: [],- }),- },- orchestrationEngine: { dispatch, readEvents: () => Stream.empty },- },- });-- const { body: tokenBody } = yield* exchangeAccessToken(defaultDesktopBootstrapToken, {- scope: "orchestration:operate",- });-- const response = yield* HttpClient.post("/api/orchestration/dispatch", {- headers: { authorization: `Bearer ${tokenBody.access_token ?? ""}` },- body: yield* HttpBody.json({- type: "thread.turn.start",- commandId: CommandId.make("cmd-http-bootstrap-capped"),- threadId: ThreadId.make("thread-http-bootstrap-capped"),- respondingBotId: BotId.make("bot-http-unrelated-responder"),- message: {- messageId: MessageId.make("msg-http-bootstrap-capped"),- role: "user",- text: "Start chat",- attachments: [],- },- runtimeMode: "full-access",- interactionMode: "default",- bootstrap: {- createThread: {- projectId: defaultProjectId,- botId,- title: "Capped bot chat",- modelSelection: defaultModelSelection,- runtimeMode: "full-access",- interactionMode: "default",- branch: "main",- worktreePath: null,- createdAt: "2026-01-01T00:00:00.000Z",- },- },- createdAt: "2026-01-01T00:00:00.000Z",- }),- });-- assert.equal(response.status, 400);- const error = (yield* response.json) as Schema.JsonObject;- assert.equal(error.unavailability, "usage-cap");- assert.equal(error.repairAction, "usage");- assert.equal(dispatch.mock.calls.length, 0);- }).pipe(Effect.provide(NodeHttpServer.layerTest)),- );-- it.effect("checks a new bot chat's usage cap before creating its thread", () =>- Effect.gen(function* () {- const botId = BotId.make("bot-bootstrap-capped");-- const bot = {- ...makeChannelTestBot(),- botId,- imageProvider: null,- usageCap: { unit: "tokens" as const, limit: 100 },- } satisfies ProjectionBots.ProjectionBot;-- const dispatch = vi.fn<OrchestrationEngine.OrchestrationEngineService["Service"]["dispatch"]>(- () => Effect.succeed({ sequence: 1 }),- );-- yield* buildAppUnderTest({- layers: {- providerRegistry: { getProviders: Effect.succeed([readyDefaultProvider]) },- projectionBots: {- getById: ({ botId: requested }) =>- Effect.succeed(requested === botId ? Option.some(bot) : Option.none()),- },- botUsageLedger: {- summarize: () =>- Effect.succeed({- botId,- consumedTokens: 100,- reservedTokens: 0,- measurements: {- input: { tokens: 100, unavailableEntries: 0 },- output: { tokens: 0, unavailableEntries: 0 },- observer: { tokens: 0, unavailableEntries: 0 },- reflector: { tokens: 0, unavailableEntries: 0 },- },- entries: [],- }),- },- orchestrationEngine: { dispatch, readEvents: () => Stream.empty },- },- });-- const wsUrl = yield* getWsServerUrl("/ws");-- const error = yield* Effect.scoped(- withWsRpcClient(wsUrl, (client) =>- client[ORCHESTRATION_WS_METHODS.dispatchCommand]({- type: "thread.turn.start",- commandId: CommandId.make("cmd-bootstrap-capped-bot"),- threadId: ThreadId.make("thread-bootstrap-capped-bot"),- respondingBotId: BotId.make("bot-unrelated-responder"),- message: {- messageId: MessageId.make("msg-bootstrap-capped-bot"),- role: "user",- text: "Start chat",- attachments: [],- },- runtimeMode: "full-access",- interactionMode: "default",- bootstrap: {- createThread: {- projectId: defaultProjectId,- botId,- title: "Capped bot chat",- modelSelection: defaultModelSelection,- runtimeMode: "full-access",- interactionMode: "default",- branch: "main",- worktreePath: null,- createdAt: "2026-01-01T00:00:00.000Z",- },- },- createdAt: "2026-01-01T00:00:00.000Z",- }).pipe(Effect.flip),- ),- );-- assert.equal(error._tag, "OrchestrationDispatchCommandError");-- if (Predicate.isTagged(error, "OrchestrationDispatchCommandError")) {- assert.equal(error.unavailability, "usage-cap");- }-- assert.equal(dispatch.mock.calls.length, 0);- }).pipe(Effect.provide(NodeHttpServer.layerTest)),- );});diff --git a/apps/server/src/server.botServices.test.ts b/apps/server/src/server.botServices.test.tsindex 0ff16d8de..babf2082a 100644--- a/apps/server/src/server.botServices.test.ts+++ b/apps/server/src/server.botServices.test.ts@@ -114,7 +114,6 @@ it.layer(NodeServices.layer)("server router seam", (it) => {engine: { provider: "codex-work", model: "gpt-5.6-sol" },sandbox: "local",runtimeMode: "approval-required" as const,- usageCap: { unit: "tokens" as const, limit: 64_000 },imageProvider: null,voiceEnabled: false,groupId: null,@@ -152,7 +151,6 @@ it.layer(NodeServices.layer)("server router seam", (it) => {),);- assert.deepEqual(result.usage.usageCap, { unit: "tokens", limit: 64_000 });assert.equal(result.usage.consumedTokens, 1_500);assert.equal(result.usage.reservedTokens, 32_000);assert.deepEqual(result.usage.measurements, {@@ -186,7 +184,6 @@ it.layer(NodeServices.layer)("server router seam", (it) => {engine: { provider: "grok", model: "grok-4.6" },sandbox: null,runtimeMode: "approval-required" as const,- usageCap: null,imageProvider: null,voiceEnabled: false,groupId: null,diff --git a/apps/server/src/server.groupAdmission.test.ts b/apps/server/src/server.groupAdmission.test.tsindex 05881dfde..faa346254 100644--- a/apps/server/src/server.groupAdmission.test.ts+++ b/apps/server/src/server.groupAdmission.test.ts@@ -1,4 +1,3 @@-import * as Schema from "effect/Schema";import * as Predicate from "effect/Predicate";import * as NodeHttpServer from "@effect/platform-node/NodeHttpServer";import * as NodeServices from "@effect/platform-node/NodeServices";@@ -20,7 +19,6 @@ import { HttpBody, HttpClient } from "effect/unstable/http";import { vi } from "vite-plus/test";import * as OrchestrationEngine from "./orchestration/Services/OrchestrationEngine.ts";import * as ProjectionBots from "./persistence/Services/ProjectionBots.ts";-import * as ProjectionGroups from "./persistence/Services/ProjectionGroups.ts";import {makeChannelTestBot,@@ -313,162 +311,6 @@ it.layer(NodeServices.layer)("server router seam", (it) => {}).pipe(Effect.provide(NodeHttpServer.layerTest)),);- it.effect("rejects capped group responders selected by boss or mention", () =>- Effect.gen(function* () {- const groupId = GroupId.make("group-usage-cap");- const bossId = BotId.make("bot-capped-boss");- const memberId = BotId.make("bot-capped-member");-- const bots = [- { ...makeChannelTestBot(), botId: bossId, name: "Boss", groupId },- { ...makeChannelTestBot(), botId: memberId, name: "Member", groupId },- ].map((bot) => ({- ...bot,- imageProvider: null,- usageCap: { unit: "tokens" as const, limit: 100 },- })) satisfies ProjectionBots.ProjectionBot[];-- const group = {- groupId,- name: "Team",- bossBotId: bossId,- members: [- { kind: "bot" as const, botId: bossId, role: "boss" as const },- { kind: "bot" as const, botId: memberId, role: "specialist" as const },- ],- createdAt: "2026-01-01T00:00:00.000Z",- updatedAt: "2026-01-01T00:00:00.000Z",- } satisfies ProjectionGroups.ProjectionGroup;-- let memberCapped = true;-- const dispatch = vi.fn<OrchestrationEngine.OrchestrationEngineService["Service"]["dispatch"]>(- () => Effect.succeed({ sequence: 1 }),- );-- yield* buildAppUnderTest({- layers: {- providerRegistry: { getProviders: Effect.succeed([readyDefaultProvider]) },- projectionSnapshotQuery: {- getThreadShellById: () =>- Effect.succeed(- Option.some(makeDefaultOrchestrationThreadShell({ groupId, botId: null })),- ),- },- projectionGroups: { getById: () => Effect.succeed(Option.some(group)) },- projectionBots: {- getById: ({ botId }) => {- const bot = bots.find((candidate) => candidate.botId === botId);-- return Effect.succeed(bot ? Option.some(bot) : Option.none());- },- listAll: () => Effect.succeed(bots),- },- botUsageLedger: {- summarize: (botId) =>- Effect.succeed({- botId,- consumedTokens: botId === memberId && !memberCapped ? 0 : 100,- reservedTokens: 0,- measurements: {- input: { tokens: 100, unavailableEntries: 0 },- output: { tokens: 0, unavailableEntries: 0 },- observer: { tokens: 0, unavailableEntries: 0 },- reflector: { tokens: 0, unavailableEntries: 0 },- },- entries: [],- }),- },- orchestrationEngine: { dispatch, readEvents: () => Stream.empty },- },- });-- const wsUrl = yield* getWsServerUrl("/ws");-- for (const [index, text] of ["Ask the team", `Ask @bot:${memberId} now`].entries()) {- const error = yield* Effect.scoped(- withWsRpcClient(wsUrl, (client) =>- client[ORCHESTRATION_WS_METHODS.dispatchCommand]({- type: "thread.turn.start",- commandId: CommandId.make(`cmd-group-cap-${index}`),- threadId: defaultThreadId,- message: {- messageId: MessageId.make(`msg-group-cap-${index}`),- role: "user",- text,- attachments: [],- },- modelSelection: defaultModelSelection,- runtimeMode: "full-access",- interactionMode: "default",- createdAt: "2026-01-01T00:00:00.000Z",- }).pipe(Effect.flip),- ),- );-- assert.equal(error._tag, "OrchestrationDispatchCommandError");-- if (Predicate.isTagged(error, "OrchestrationDispatchCommandError")) {- assert.equal(error.unavailability, "usage-cap");- }- }-- assert.equal(dispatch.mock.calls.length, 0);-- const { body: tokenBody } = yield* exchangeAccessToken(defaultDesktopBootstrapToken, {- scope: "orchestration:operate",- });-- for (const [index, text] of ["Ask the team", `Ask @bot:${memberId} now`].entries()) {- const response = yield* HttpClient.post("/api/orchestration/dispatch", {- headers: { authorization: `Bearer ${tokenBody.access_token ?? ""}` },- body: yield* HttpBody.json({- type: "thread.turn.start",- commandId: CommandId.make(`cmd-http-group-cap-${index}`),- threadId: defaultThreadId,- message: {- messageId: MessageId.make(`msg-http-group-cap-${index}`),- role: "user",- text,- attachments: [],- },- modelSelection: defaultModelSelection,- runtimeMode: "full-access",- interactionMode: "default",- createdAt: "2026-01-01T00:00:00.000Z",- }),- });-- assert.equal(response.status, 400);- const error = (yield* response.json) as Schema.JsonObject;- assert.equal(error.unavailability, "usage-cap");- }-- assert.equal(dispatch.mock.calls.length, 0);- memberCapped = false;- yield* Effect.scoped(- withWsRpcClient(wsUrl, (client) =>- client[ORCHESTRATION_WS_METHODS.dispatchCommand]({- type: "thread.turn.start",- commandId: CommandId.make("cmd-group-cap-uncapped"),- threadId: defaultThreadId,- message: {- messageId: MessageId.make("msg-group-cap-uncapped"),- role: "user",- text: `Ask @bot:${memberId} now`,- attachments: [],- },- modelSelection: defaultModelSelection,- runtimeMode: "full-access",- interactionMode: "default",- createdAt: "2026-01-01T00:00:00.000Z",- }),- ),- );- assert.equal(dispatch.mock.calls.length, 1);- }).pipe(Effect.provide(NodeHttpServer.layerTest)),- );-it.effect("retries a turn after a transient provider failure", () =>Effect.gen(function* () {const dispatch = vi.fn<OrchestrationEngine.OrchestrationEngineService["Service"]["dispatch"]>(diff --git a/apps/server/src/server.modelAdmission.test.ts b/apps/server/src/server.modelAdmission.test.tsindex 3127effc7..a0f1e3319 100644--- a/apps/server/src/server.modelAdmission.test.ts+++ b/apps/server/src/server.modelAdmission.test.ts@@ -93,7 +93,6 @@ it.describe("ws bot engine model routing preflight", () => {engine,sandbox: "local" as const,runtimeMode: "full-access" as const,- usageCap: null,imageProvider: null,voiceEnabled: false,groupId: null,@@ -150,7 +149,6 @@ it.describe("ws bot engine model routing preflight", () => {avatar: { kind: "dither", seed: "bad-model" },engine: { provider: "codex", model: "not-a-model" },sandbox: "local",- usageCap: null,groupId: null,createdAt: now,}),@@ -204,7 +202,6 @@ it.describe("ws bot engine model routing preflight", () => {avatar: { kind: "dither", seed: "bad-model" },engine: { provider: "codex", model: "not-a-model" },sandbox: "local",- usageCap: null,groupId: null,createdAt: now,}),diff --git a/apps/server/src/serverRuntimeStartup.delegations.test.ts b/apps/server/src/serverRuntimeStartup.delegations.test.tsindex 32bcad467..af85b8a5b 100644--- a/apps/server/src/serverRuntimeStartup.delegations.test.ts+++ b/apps/server/src/serverRuntimeStartup.delegations.test.ts@@ -177,7 +177,6 @@ const seed = Effect.gen(function* () {engine: null,sandbox: "local",runtimeMode: "approval-required",- usageCap: null,groupId: null,createdAt: CREATED_AT,});diff --git a/apps/server/src/serverTestFixtures.ts b/apps/server/src/serverTestFixtures.tsindex 9533e11f7..d84d7ebd6 100644--- a/apps/server/src/serverTestFixtures.ts+++ b/apps/server/src/serverTestFixtures.ts@@ -157,7 +157,6 @@ export const makeChannelTestBot = () => ({engine: null,sandbox: "local" as const,runtimeMode: "full-access" as const,- usageCap: null,voiceEnabled: false,imageProvider: null,channelBindings: [],diff --git a/apps/server/src/usage/BotUsageLedger.recovery.test.ts b/apps/server/src/usage/BotUsageLedger.recovery.test.tsindex 9c1c70ba6..808b0f5eb 100644--- a/apps/server/src/usage/BotUsageLedger.recovery.test.ts+++ b/apps/server/src/usage/BotUsageLedger.recovery.test.ts@@ -29,7 +29,6 @@ it.effect("reconciles persisted reservations when the ledger restarts", () =>botId,threadId,maximumTokens: 100,- capLimit: 1_000,}),);yield* ledger.reserve(@@ -38,7 +37,6 @@ it.effect("reconciles persisted reservations when the ledger restarts", () =>threadId,turnId: TurnId.make("turn-interrupted"),maximumTokens: 200,- capLimit: 1_000,}),);yield* ledger.reserve(@@ -47,15 +45,10 @@ it.effect("reconciles persisted reservations when the ledger restarts", () =>threadId,turnId: TurnId.make("turn-reported"),maximumTokens: 300,- capLimit: 1_000,}),);- const {- maximumTokens: _maximumTokens,- capLimit: _capLimit,- ...toolStart- } = reserveInput("tool-before-restart", {+ const { maximumTokens: _maximumTokens, ...toolStart } = reserveInput("tool-before-restart", {botId,threadId,turnId: TurnId.make("turn-interrupted"),@@ -163,11 +156,12 @@ it.effect("keeps pricing complete when a restart interrupts a tool call", () =>settledAt: "2026-08-30T20:02:00.000Z",});- const {- maximumTokens: _maximumTokens,- capLimit: _capLimit,- ...toolStart- } = reserveInput("interrupted-tool", { botId, threadId, turnId, category: "tool" });+ const { maximumTokens: _maximumTokens, ...toolStart } = reserveInput("interrupted-tool", {+ botId,+ threadId,+ turnId,+ category: "tool",+ });yield* ledger.recordStart(toolStart);}).pipe(Effect.provide(restartedLayer()));diff --git a/apps/server/src/usage/BotUsageLedger.settlement.test.ts b/apps/server/src/usage/BotUsageLedger.settlement.test.tsindex 63b88f5b1..3c20ca94c 100644--- a/apps/server/src/usage/BotUsageLedger.settlement.test.ts+++ b/apps/server/src/usage/BotUsageLedger.settlement.test.ts@@ -66,7 +66,7 @@ it.layer(layer)("BotUsageLedger", (it) => {}),);- it.effect("records tool and routine writers and includes their priced tokens in the cap", () =>+ it.effect("records tool and routine writers and prices their tokens", () =>Effect.gen(function* () {const ledger = yield* BotUsageLedger;const botId = BotId.make("bot-tool-routine");@@ -76,7 +76,6 @@ it.layer(layer)("BotUsageLedger", (it) => {botId,category: "tool",maximumTokens: 100,- capLimit: 100,}),);@@ -120,18 +119,6 @@ it.layer(layer)("BotUsageLedger", (it) => {);assert.equal(cost.costUsd, 90);-- const remaining = yield* ledger.reserve(- reserveInput("cap-after-tool", { botId, maximumTokens: 41, capLimit: 100 }),- );-- assert.equal(remaining.reservedTokens, 40);-- const rejected = yield* ledger- .reserve(reserveInput("cap-after-tool-2", { botId, maximumTokens: 1, capLimit: 100 }))- .pipe(Effect.exit);-- assert.equal(rejected._tag, "Failure");}),);@@ -150,7 +137,6 @@ it.layer(layer)("BotUsageLedger", (it) => {turnId,category: "observer",maximumTokens: 32_000,- capLimit: 32_000,}),);@@ -209,7 +195,6 @@ it.layer(layer)("BotUsageLedger", (it) => {provider: ProviderDriverKind.make(provider),model,maximumTokens: 32_000,- capLimit: 32_000,}),);diff --git a/apps/server/src/usage/BotUsageLedger.test.ts b/apps/server/src/usage/BotUsageLedger.test.tsindex 15448996e..760de6921 100644--- a/apps/server/src/usage/BotUsageLedger.test.ts+++ b/apps/server/src/usage/BotUsageLedger.test.ts@@ -1,4 +1,3 @@-import * as Predicate from "effect/Predicate";import { layer, reserveInput } from "./testUtils/botUsageLedger.ts";import { assert, it } from "@effect/vitest";import {@@ -60,7 +59,6 @@ it.layer(layer)("BotUsageLedger", (it) => {sourceKey,botId: firstBotId,maximumTokens: 60,- capLimit: 100,}),);yield* ledger.reserve(@@ -68,7 +66,6 @@ it.layer(layer)("BotUsageLedger", (it) => {sourceKey,botId: secondBotId,maximumTokens: 60,- capLimit: 100,}),);@@ -81,57 +78,13 @@ it.layer(layer)("BotUsageLedger", (it) => {}),);- it.effect("rejects a second reservation when a cap is fully reserved", () =>- Effect.gen(function* () {- const ledger = yield* BotUsageLedger;- const botId = BotId.make("bot-cap");- yield* ledger.reserve(- reserveInput("cap-first", { botId, maximumTokens: 100, capLimit: 100 }),- );-- const exit = yield* ledger- .reserve(reserveInput("cap-second", { botId, maximumTokens: 100, capLimit: 100 }))- .pipe(Effect.exit);-- assert.isTrue(Predicate.isTagged(exit, "Failure"));- const summary = yield* ledger.summarize(botId);- assert.equal(summary.reservedTokens, 100);- assert.equal(summary.entries.length, 1);- }),- );-- it.effect("reserves the remaining cap when a request exceeds it", () =>- Effect.gen(function* () {- const ledger = yield* BotUsageLedger;- const botId = BotId.make("bot-partial-cap");- yield* ledger.reserve(- reserveInput("partial-first", { botId, maximumTokens: 60, capLimit: 100 }),- );-- const reservation = yield* ledger.reserve(- reserveInput("partial-second", { botId, maximumTokens: 50, capLimit: 100 }),- );-- assert.equal(reservation.reservedTokens, 40);-- const failure = yield* ledger- .reserve(reserveInput("partial-third", { botId, maximumTokens: 1, capLimit: 100 }))- .pipe(Effect.flip);-- assert.equal(failure._tag, "BotUsageCapExceeded");- const summary = yield* ledger.summarize(botId);- assert.equal(summary.reservedTokens, 100);- assert.equal(summary.entries.length, 2);- }),- );-it.effect("charges the reservation when provider usage is unavailable", () =>Effect.gen(function* () {const ledger = yield* BotUsageLedger;const botId = BotId.make("bot-unavailable");const reservation = yield* ledger.reserve(- reserveInput("unavailable", { botId, maximumTokens: 500, capLimit: 500 }),+ reserveInput("unavailable", { botId, maximumTokens: 500 }),);yield* ledger.settle({@@ -161,9 +114,7 @@ it.layer(layer)("BotUsageLedger", (it) => {Effect.gen(function* () {const ledger = yield* BotUsageLedger;const botId = BotId.make("bot-cancelled");- yield* ledger.reserve(- reserveInput("cancelled", { botId, maximumTokens: 500, capLimit: 500 }),- );+ yield* ledger.reserve(reserveInput("cancelled", { botId, maximumTokens: 500 }));yield* ledger.finalizeForTurn({botId,threadId: ThreadId.make("thread-1"),@@ -184,9 +135,7 @@ it.layer(layer)("BotUsageLedger", (it) => {const botId = BotId.make("bot-late-cancellation");const threadId = ThreadId.make("thread-1");const oldTurnId = TurnId.make("turn-old");- yield* ledger.reserve(- reserveInput("old-turn", { botId, threadId, maximumTokens: 500, capLimit: 1_000 }),- );+ yield* ledger.reserve(reserveInput("old-turn", { botId, threadId, maximumTokens: 500 }));yield* ledger.finalizeForTurn({botId,threadId,@@ -194,9 +143,7 @@ it.layer(layer)("BotUsageLedger", (it) => {settledAt: "2026-08-30T20:01:00.000Z",cancelled: true,});- yield* ledger.reserve(- reserveInput("new-turn", { botId, threadId, maximumTokens: 500, capLimit: 1_000 }),- );+ yield* ledger.reserve(reserveInput("new-turn", { botId, threadId, maximumTokens: 500 }));yield* ledger.finalizeForTurn({botId,threadId,@@ -223,9 +170,7 @@ it.layer(layer)("BotUsageLedger", (it) => {const ledger = yield* BotUsageLedger;const botId = BotId.make("bot-cancelled-after-report");const turnId = TurnId.make("turn-cancelled-after-report");- yield* ledger.reserve(- reserveInput("cancelled-after-report", { botId, maximumTokens: 500, capLimit: 500 }),- );+ yield* ledger.reserve(reserveInput("cancelled-after-report", { botId, maximumTokens: 500 }));yield* ledger.settleForTurn({botId,threadId: ThreadId.make("thread-1"),@@ -258,7 +203,7 @@ it.layer(layer)("BotUsageLedger", (it) => {const botId = BotId.make("bot-reported-overage");const reservation = yield* ledger.reserve(- reserveInput("reported-overage", { botId, maximumTokens: 100, capLimit: 100 }),+ reserveInput("reported-overage", { botId, maximumTokens: 100 }),);yield* ledger.settle({@@ -322,7 +267,6 @@ it.layer(layer)("BotUsageLedger", (it) => {sourceKey: reservationId,category,maximumTokens: 100,- capLimit: 1_000,}),);@@ -354,7 +298,6 @@ it.layer(layer)("BotUsageLedger", (it) => {sourceKey: "delegate:chief:research",category: "delegated",maximumTokens: 200,- capLimit: 1_000,}),);diff --git a/apps/server/src/usage/BotUsageLedger.ts b/apps/server/src/usage/BotUsageLedger.tsindex a2acd23e7..890f77e26 100644--- a/apps/server/src/usage/BotUsageLedger.ts+++ b/apps/server/src/usage/BotUsageLedger.ts@@ -13,11 +13,7 @@ import {toPersistenceDecodeError,toPersistenceSqlError,} from "../persistence/Errors.ts";-import {- BotUsageCapExceeded,- type SettleBotUsageDetails,- type BotUsageLedgerShape,-} from "./BotUsageLedgerTypes.ts";+import { type SettleBotUsageDetails, type BotUsageLedgerShape } from "./BotUsageLedgerTypes.ts";import {UsageRow,entryColumns,@@ -156,7 +152,7 @@ const make = Effect.gen(function* () {sql.withTransaction(Effect.gen(function* () {- yield* validateTokens("BotUsageLedger.reserve", [input.maximumTokens, input.capLimit]);+ yield* validateTokens("BotUsageLedger.reserve", [input.maximumTokens]);const prior = yield* selectEntryBySource(sql, input.botId, input.sourceKey);if (prior[0]) return yield* decodeEntry(prior[0]);@@ -167,31 +163,9 @@ const make = Effect.gen(function* () {ON CONFLICT (bot_id) DO NOTHING`;- const balance = yield* sql<{- readonly consumedTokens: number;- readonly reservedTokens: number;- }>`- SELECT consumed_tokens AS "consumedTokens", reserved_tokens AS "reservedTokens"- FROM akeru_bot_usage_balances WHERE bot_id = ${input.botId}- `;-- const current = balance[0]!;- const available = input.capLimit - current.consumedTokens - current.reservedTokens;-- if (input.maximumTokens <= 0 || available <= 0) {- return yield* new BotUsageCapExceeded({- botId: input.botId,- limit: input.capLimit,- consumedTokens: current.consumedTokens,- reservedTokens: current.reservedTokens,- requestedTokens: input.maximumTokens,- });- }-- const reservedTokens = Math.min(input.maximumTokens, available);yield* sql`UPDATE akeru_bot_usage_balances- SET reserved_tokens = reserved_tokens + ${reservedTokens}, updated_at = ${input.createdAt}+ SET reserved_tokens = reserved_tokens + ${input.maximumTokens}, updated_at = ${input.createdAt}WHERE bot_id = ${input.botId}`;yield* sql`@@ -201,7 +175,7 @@ const make = Effect.gen(function* () {model, unavailable_reason, created_at, settled_at) VALUES (${input.reservationId}, ${input.sourceKey}, ${input.botId}, ${input.threadId},- ${input.turnId}, ${input.category}, 'reserved', ${reservedTokens}, ${reservedTokens},+ ${input.turnId}, ${input.category}, 'reserved', ${input.maximumTokens}, ${input.maximumTokens},NULL, NULL, NULL, ${input.provider}, ${input.model}, NULL, ${input.createdAt}, NULL)`;@@ -212,11 +186,9 @@ const make = Effect.gen(function* () {).pipe(Effect.mapError((cause) =>- Predicate.isTagged(cause, "BotUsageCapExceeded")+ Predicate.isTagged(cause, "PersistenceDecodeError")? cause- : Predicate.isTagged(cause, "PersistenceDecodeError")- ? cause- : toPersistenceSqlError("BotUsageLedger.reserve")(cause),+ : toPersistenceSqlError("BotUsageLedger.reserve")(cause),),),);@@ -706,8 +678,6 @@ const decodeUsageSummary = Schema.decodeUnknownEffect(AkeruBotUsageSummary);export const BotUsageLedgerLive = Layer.effect(BotUsageLedger, make);-export { BotUsageCapExceeded } from "./BotUsageLedgerTypes.ts";-export { AKERU_TURN_USAGE_RESERVATION_TOKENS } from "./BotUsageLedgerTypes.ts";export type { ReserveBotUsageInput } from "./BotUsageLedgerTypes.ts";diff --git a/apps/server/src/usage/BotUsageLedgerTypes.ts b/apps/server/src/usage/BotUsageLedgerTypes.tsindex 08dc0dfd6..36bc2c694 100644--- a/apps/server/src/usage/BotUsageLedgerTypes.ts+++ b/apps/server/src/usage/BotUsageLedgerTypes.ts@@ -9,24 +9,8 @@ import {type TurnId,} from "@akeru/contracts";import * as Effect from "effect/Effect";-import * as Schema from "effect/Schema";import { type PersistenceDecodeError, PersistenceSqlError } from "../persistence/Errors.ts";-export class BotUsageCapExceeded extends Schema.TaggedErrorClass<BotUsageCapExceeded>()(- "BotUsageCapExceeded",- {- botId: Schema.String,- limit: Schema.Number,- consumedTokens: Schema.Number,- reservedTokens: Schema.Number,- requestedTokens: Schema.Number,- },-) {- override get message(): string {- return `Bot ${this.botId} reached its ${this.limit}-token usage cap.`;- }-}-export const AKERU_TURN_USAGE_RESERVATION_TOKENS = 32_000;export interface ReserveBotUsageInput {@@ -37,7 +21,6 @@ export interface ReserveBotUsageInput {readonly turnId: TurnId | null;readonly category: AkeruUsageCategory;readonly maximumTokens: number;- readonly capLimit: number;readonly provider: ProviderDriverKind | null;readonly model: string | null;readonly createdAt: string;@@ -67,10 +50,7 @@ export type SettleBotUsageForTurnInput = SettleBotUsageDetails & {readonly settledAt: string;};-export type BotUsageLedgerError =- | BotUsageCapExceeded- | PersistenceSqlError- | PersistenceDecodeError;+export type BotUsageLedgerError = PersistenceSqlError | PersistenceDecodeError;export interface BotUsageLedgerShape {readonly pricingTotals: (botId: BotId) => Effect.Effect<@@ -85,34 +65,28 @@ export interface BotUsageLedgerShape {readonly reasoningTokens: number;}>;},- Exclude<BotUsageLedgerError, BotUsageCapExceeded>+ BotUsageLedgerError>;readonly reserve: (input: ReserveBotUsageInput,) => Effect.Effect<AkeruUsageEntry, BotUsageLedgerError>;readonly settle: (input: SettleBotUsageInput,- ) => Effect.Effect<AkeruUsageEntry, Exclude<BotUsageLedgerError, BotUsageCapExceeded>>;+ ) => Effect.Effect<AkeruUsageEntry, BotUsageLedgerError>;readonly bindTurn: (input: {readonly reservationId: AkeruUsageReservationId;readonly turnId: TurnId;- }) => Effect.Effect<AkeruUsageEntry, Exclude<BotUsageLedgerError, BotUsageCapExceeded>>;+ }) => Effect.Effect<AkeruUsageEntry, BotUsageLedgerError>;readonly settleForTurn: (input: SettleBotUsageForTurnInput,- ) => Effect.Effect<- ReadonlyArray<AkeruUsageEntry>,- Exclude<BotUsageLedgerError, BotUsageCapExceeded>- >;+ ) => Effect.Effect<ReadonlyArray<AkeruUsageEntry>, BotUsageLedgerError>;readonly finalizeForTurn: (input: {readonly botId: BotId;readonly threadId: ThreadId;readonly turnId: TurnId;readonly settledAt: string;readonly cancelled?: boolean;- }) => Effect.Effect<- ReadonlyArray<AkeruUsageEntry>,- Exclude<BotUsageLedgerError, BotUsageCapExceeded>- >;+ }) => Effect.Effect<ReadonlyArray<AkeruUsageEntry>, BotUsageLedgerError>;readonly recordMeasurement: (input: {readonly reservationId: AkeruUsageReservationId;readonly sourceKey: string;@@ -129,12 +103,10 @@ export interface BotUsageLedgerShape {readonly model: string | null;readonly includedInReservation?: boolean;readonly createdAt: string;- }) => Effect.Effect<AkeruUsageEntry, Exclude<BotUsageLedgerError, BotUsageCapExceeded>>;+ }) => Effect.Effect<AkeruUsageEntry, BotUsageLedgerError>;/** Records work that charges no tokens so a restart before it settles marks it interrupted. */readonly recordStart: (- input: Omit<ReserveBotUsageInput, "maximumTokens" | "capLimit">,- ) => Effect.Effect<AkeruUsageEntry, Exclude<BotUsageLedgerError, BotUsageCapExceeded>>;- readonly summarize: (- botId: BotId,- ) => Effect.Effect<AkeruBotUsageSummary, Exclude<BotUsageLedgerError, BotUsageCapExceeded>>;+ input: Omit<ReserveBotUsageInput, "maximumTokens">,+ ) => Effect.Effect<AkeruUsageEntry, BotUsageLedgerError>;+ readonly summarize: (botId: BotId) => Effect.Effect<AkeruBotUsageSummary, BotUsageLedgerError>;}diff --git a/apps/server/src/usage/testUtils/botUsageLedger.ts b/apps/server/src/usage/testUtils/botUsageLedger.tsindex 4806f7508..89c74cc18 100644--- a/apps/server/src/usage/testUtils/botUsageLedger.ts+++ b/apps/server/src/usage/testUtils/botUsageLedger.ts@@ -16,7 +16,6 @@ const reserveInput = (turnId: null,category: "turn",maximumTokens: 1_000,- capLimit: 1_000,provider: ProviderDriverKind.make("codex"),model: "gpt-5.6-sol",createdAt: "2026-08-30T20:00:00.000Z",diff --git a/apps/server/src/voiceCall/VoiceCallManager.test.ts b/apps/server/src/voiceCall/VoiceCallManager.test.tsindex bbf5b955e..8778f0b77 100644--- a/apps/server/src/voiceCall/VoiceCallManager.test.ts+++ b/apps/server/src/voiceCall/VoiceCallManager.test.ts@@ -110,7 +110,6 @@ const bot = {engine: null,sandbox: null,runtimeMode: "full-access" as const,- usageCap: null,imageProvider: null,voiceEnabled: true,groupId: null,diff --git a/apps/server/src/voiceCall/testUtils/voiceProviders.ts b/apps/server/src/voiceCall/testUtils/voiceProviders.tsindex e6f6fd9ba..8bcd25975 100644--- a/apps/server/src/voiceCall/testUtils/voiceProviders.ts+++ b/apps/server/src/voiceCall/testUtils/voiceProviders.ts@@ -31,7 +31,6 @@ const repository = {engine: null,sandbox: null,runtimeMode: "full-access" as const,- usageCap: null,imageProvider: null,voiceEnabled: true,groupId: null,diff --git a/apps/server/src/wsBotServicesHandlers.ts b/apps/server/src/wsBotServicesHandlers.tsindex a11529e3c..522620ad1 100644--- a/apps/server/src/wsBotServicesHandlers.ts+++ b/apps/server/src/wsBotServicesHandlers.ts@@ -198,7 +198,6 @@ export const createWsBotServicesHandlers = ({return {...summary,- usageCap: bot.value.usageCap,estimatedCost,subscriptionPool,};diff --git a/apps/server/src/wsOrchestrationHandlers.ts b/apps/server/src/wsOrchestrationHandlers.tsindex cda6b3bcc..21d44e7d2 100644--- a/apps/server/src/wsOrchestrationHandlers.ts+++ b/apps/server/src/wsOrchestrationHandlers.ts@@ -37,7 +37,6 @@ export const createWsOrchestrationHandlers = ({currentSessionId,projectionSnapshotQuery,projectionBots,- botUsageLedger,projectionGroups,checkpointDiffQuery,providerRegistry,@@ -57,7 +56,6 @@ export const createWsOrchestrationHandlers = ({| "currentSessionId"| "projectionSnapshotQuery"| "projectionBots"- | "botUsageLedger"| "projectionGroups"| "checkpointDiffQuery"| "providerRegistry"@@ -282,22 +280,6 @@ export const createWsOrchestrationHandlers = ({});}}-- if (botId) {- if (bot?.usageCap) {- const usage = yield* botUsageLedger.summarize(botId);-- if (usage.consumedTokens + usage.reservedTokens >= bot.usageCap.limit) {- yield* cleanupFailedUploadedAttachments(command, normalizedCommand);-- return yield* new OrchestrationDispatchCommandError({- message: `Usage cap reached for ${bot.name}.`,- unavailability: "usage-cap",- repairAction: "usage",- });- }- }- }}// Archive and settle both mean "done with this thread", so adiff --git a/apps/server/src/wsOrchestrationStreams.ts b/apps/server/src/wsOrchestrationStreams.tsindex 2644afb82..c16afe73f 100644--- a/apps/server/src/wsOrchestrationStreams.ts+++ b/apps/server/src/wsOrchestrationStreams.ts@@ -251,7 +251,6 @@ export const createWsOrchestrationStreams = ({engine: nextBot.engine,sandbox: nextBot.sandbox,runtimeMode: nextBot.runtimeMode,- usageCap: nextBot.usageCap,imageProvider: nextBot.imageProvider,personalityTone: nextBot.personalityTone ?? BALANCED_BOT_PERSONALITY_TONE,voiceEnabled: nextBot.voiceEnabled,diff --git a/apps/web/src/components/onboarding/DesktopOnboardingSurface.tsx b/apps/web/src/components/onboarding/DesktopOnboardingSurface.tsxindex 840d27fff..98b439007 100644--- a/apps/web/src/components/onboarding/DesktopOnboardingSurface.tsx+++ b/apps/web/src/components/onboarding/DesktopOnboardingSurface.tsx@@ -258,7 +258,6 @@ export function OnboardingSurface({engine: providerReadiness.engine,sandbox: null,runtimeMode: DEFAULT_BOT_RUNTIME_MODE,- usageCap: null,groupId: null,},});diff --git a/apps/web/src/components/roster/BotChatMessageRows.test.ts b/apps/web/src/components/roster/BotChatMessageRows.test.tsindex 1a6911dff..2d5770345 100644--- a/apps/web/src/components/roster/BotChatMessageRows.test.ts+++ b/apps/web/src/components/roster/BotChatMessageRows.test.ts@@ -28,7 +28,7 @@ function props(overrides: Partial<RowProps> = {}): RowProps {testId: "row",cwd: "/work",threadRef: { environmentId: EnvironmentId.make("env"), threadId: ThreadId.make("thread") },- stepMeter: { engine, tokens: 10, costUsd: null, hardStopReached: false },+ stepMeter: { engine, tokens: 10, costUsd: null },pluginResults: [{ id: "entry-1", result }],currentPersonId: "person",playback: null,diff --git a/apps/web/src/components/roster/BotComposerModelControl.test.tsx b/apps/web/src/components/roster/BotComposerModelControl.test.tsxnew file mode 100644index 000000000..13ccd781b--- /dev/null+++ b/apps/web/src/components/roster/BotComposerModelControl.test.tsx@@ -0,0 +1,239 @@+import type { TestValue } from "../test-support/reactTree";+import { decodeServerProvider } from "../test-support/fixtures";+import { Predicate } from "effect";+import {+ DEFAULT_UNIFIED_SETTINGS,+ EnvironmentId,+ ProviderDriverKind,+ ProviderInstanceId,+ type ModelCapabilities,+ type ServerProvider,+} from "@akeru/contracts";+import { beforeEach, describe, expect, it, vi } from "vite-plus/test";++import { visitElements } from "../test-support/reactTree";+import { reactHookHarness as hooks } from "../../test/reactHookHarness";+import type { Bot } from "./types";++const atoms = vi.hoisted(() => ({+ providers: Symbol("providers"),+ keybindings: Symbol("keybindings"),+ update: Symbol("botUpdate"),+}));++const state = vi.hoisted(() => ({+ bots: [] as Bot[],+ providers: [] as ServerProvider[],+ updateBot: vi.fn(),+}));++vi.mock("react", async (importOriginal) => {+ const actual = await importOriginal<typeof import("react")>();+ const { reactHookHarness } = await import("../../test/reactHookHarness");++ return {+ ...actual,+ useCallback: reactHookHarness.useCallback,+ useEffect: () => {},+ useMemo: reactHookHarness.useMemo,+ useRef: reactHookHarness.useRef,+ useState: reactHookHarness.useState,+ };+});++vi.mock("react/compiler-runtime", async () => {+ const { reactHookHarness } = await import("../../test/reactHookHarness");++ return { c: reactHookHarness.useMemoCache };+});++vi.mock("@effect/atom-react", () => ({+ useAtomValue: (atom: symbol) => (atom === atoms.providers ? state.providers : []),+}));++vi.mock("../../state/server", () => ({+ primaryServerProvidersAtom: atoms.providers,+ primaryServerKeybindingsAtom: atoms.keybindings,+}));++vi.mock("../../state/bots", () => ({ botEnvironment: { update: atoms.update } }));++vi.mock("../../state/environments", () => ({+ usePrimaryEnvironmentId: () => EnvironmentId.make("environment-1"),+}));++vi.mock("../../state/use-atom-command", () => ({+ useAtomCommand: (atom: symbol) => (atom === atoms.update ? state.updateBot : vi.fn()),+}));++vi.mock("../../hooks/useSettings", () => ({+ usePrimarySettings: () => DEFAULT_UNIFIED_SETTINGS,+}));++vi.mock("./rosterStore", () => ({+ useRosterStore: <T,>(selector: (store: { bots: Bot[] }) => T) => selector({ bots: state.bots }),+}));++vi.mock("../ui/toast", () => ({ toastManager: { add: vi.fn() } }));++vi.mock("../../i18n", async () => {+ const { createTranslator } = await import("@akeru/client-runtime/i18n");+ const translator = createTranslator("en");++ return { useI18n: () => ({ ...translator, t: translator.translate }) };+});++import { BotComposerModelControl } from "./BotComposerModelControl";++const reasoning = (levels: ReadonlyArray<string>): ModelCapabilities => ({+ optionDescriptors: [+ {+ id: "reasoningEffort",+ label: "Reasoning",+ type: "select",+ currentValue: "default",+ options: [+ { id: "default", label: "Provider default", isDefault: true },+ ...levels.map((id) => ({ id, label: id })),+ ],+ },+ ],+});++function provider(instance: string): ServerProvider {+ return decodeServerProvider({+ instanceId: ProviderInstanceId.make(instance),+ driver: ProviderDriverKind.make("codex"),+ enabled: true,+ installed: true,+ version: "1.0.0",+ status: "ready",+ auth: { status: "authenticated" },+ checkedAt: "2026-09-01T00:00:00.000Z",+ models: [+ {+ slug: "gpt-a",+ name: "A",+ isCustom: false,+ isDefault: true,+ capabilities: reasoning(["minimal", "high"]),+ },+ { slug: "gpt-b", name: "B", isCustom: false, capabilities: reasoning(["high"]) },+ ],+ slashCommands: [],+ skills: [],+ });+}++function makeBot(overrides: Partial<Bot> = {}): Bot {+ return {+ id: "bot-1",+ name: "Scout",+ title: "Scout",+ label: null,+ description: null,+ disabledMcpServerIds: [],+ avatar: { kind: "blob", shape: "circle", color: "#2E8EFF" },+ engine: {+ provider: "codex",+ model: "gpt-a",+ options: [{ id: "reasoningEffort", value: "minimal" }],+ },+ sandbox: "local",+ runtimeMode: "full-access",+ personalityTone: 50,+ voiceEnabled: false,+ imageProvider: null,+ groupId: null,+ pinned: false,+ archivedAt: null,+ createdAt: "2026-09-01T00:00:00.000Z",+ updatedAt: "2026-09-01T00:00:00.000Z",+ ...overrides,+ };+}++function render(botId = "bot-1", disabled = false): TestValue {+ hooks.beginRender();++ return BotComposerModelControl({ botId, disabled }) as TestValue;+}++function traitsPicker(tree: TestValue) {+ return visitElements(tree, (element) => Predicate.isFunction(element.props.onModelOptionsChange));+}++function modelPicker(tree: TestValue) {+ const picker = visitElements(tree, (element) =>+ Predicate.isFunction(element.props.onInstanceModelChange),+ );++ expect(picker).not.toBeNull();++ return picker!.props as {+ readonly disabled: boolean;+ readonly onInstanceModelChange: (instanceId: ProviderInstanceId, model: string) => void;+ };+}++function savedEngine() {+ return state.updateBot.mock.calls.at(-1)?.[0];+}++describe("bot composer model control", () => {+ beforeEach(() => {+ hooks.reset();+ state.bots = [makeBot()];+ state.providers = [provider("codex"), provider("codex_work")];+ state.updateBot = vi.fn(async () => ({ _tag: "Success" }));+ });++ it("offers model selection without a composer reasoning control", () => {+ const tree = render();++ expect(traitsPicker(tree)).toBeNull();+ expect(modelPicker(tree).disabled).toBe(false);+ expect(state.updateBot).not.toHaveBeenCalled();+ });++ it("keeps supported choices across models on one instance and clears them across instances", () => {+ const picker = modelPicker(render());++ picker.onInstanceModelChange(ProviderInstanceId.make("codex"), "gpt-b");+ expect(savedEngine()?.input.engine).toEqual({ provider: "codex", model: "gpt-b" });++ state.bots = [+ makeBot({+ engine: {+ provider: "codex",+ model: "gpt-a",+ options: [{ id: "reasoningEffort", value: "high" }],+ },+ }),+ ];+ modelPicker(render()).onInstanceModelChange(ProviderInstanceId.make("codex"), "gpt-b");+ expect(savedEngine()?.input.engine).toEqual({+ provider: "codex",+ model: "gpt-b",+ options: [{ id: "reasoningEffort", value: "high" }],+ });++ modelPicker(render()).onInstanceModelChange(ProviderInstanceId.make("codex_work"), "gpt-a");+ expect(savedEngine()?.input.engine).toEqual({ provider: "codex_work", model: "gpt-a" });+ });++ it("hides the reasoning control in a read-only composer", () => {+ const tree = render("bot-1", true);++ expect(traitsPicker(tree)).toBeNull();+ expect(modelPicker(tree).disabled).toBe(true);+ });++ it("renders nothing for a group or archived bot, so no engine is written", () => {+ expect(render("group-1")).toBeNull();++ state.bots = [makeBot({ archivedAt: "2026-09-02T00:00:00.000Z" })];+ expect(render()).toBeNull();+ expect(state.updateBot).not.toHaveBeenCalled();+ });+});diff --git a/apps/web/src/components/roster/BotComposerModelControl.tsx b/apps/web/src/components/roster/BotComposerModelControl.tsxindex e691ecb87..18add9c12 100644--- a/apps/web/src/components/roster/BotComposerModelControl.tsx+++ b/apps/web/src/components/roster/BotComposerModelControl.tsx@@ -1,6 +1,6 @@import { Predicate } from "effect";import { useAtomValue } from "@effect/atom-react";-import { BotId, ProviderInstanceId } from "@akeru/contracts";+import { BotId, ProviderInstanceId, type BotEngine } from "@akeru/contracts";import { useEffect, useMemo, useState } from "react";import { usePrimarySettings } from "../../hooks/useSettings";@@ -22,13 +22,13 @@ import { primaryServerKeybindingsAtom, primaryServerProvidersAtom } from "../../import { useAtomCommand } from "../../state/use-atom-command";import { ProviderModelPicker } from "../chat/ProviderModelPicker";import { toastManager } from "../ui/toast";-import { resolveStickyBotEngine } from "./botEngineSelection";+import { botEngineForModelPick, resolveStickyBotEngine } from "./botEngineSelection";import { useRosterStore } from "./rosterStore";/*** The bot's engine at the composer: shows which model answers the next message- * and opens the shared model picker to change it. Changing the model here- * writes the bot's engine, the same field the details panel edits.+ * and opens the shared model picker. Model changes retain the bot's saved+ * reasoning when the selected model supports it.*/export function BotComposerModelControl({botId,@@ -110,26 +110,19 @@ export function BotComposerModelControl({return () => window.removeEventListener("keydown", onKeyDown);}, [keybindings, pickerOpen, selectable]);- if (selection === null) return null;+ // Only a live bot's own composer edits an engine; a group or archived bot never does.+ if (selection === null || !bot || bot.archivedAt !== null) return null;- const changeModel = async (instanceId: ProviderInstanceId, model: string) => {- if (environmentId === null || !bot) return;-- const options =- defaultSelection.instanceId === instanceId && defaultSelection.model === model- ? defaultSelection.options- : undefined;+ const saveEngine = async (engine: BotEngine, failureTitle: string) => {+ if (environmentId === null) return;const result = await updateBot({environmentId,- input: {- botId: BotId.make(bot.id),- engine: { provider: instanceId, model, ...(options ? { options } : {}) },- },+ input: { botId: BotId.make(bot.id), engine },});if (Predicate.isTagged(result, "Failure")) {- toastManager.add({ type: "error", title: t("Could not change the model") });+ toastManager.add({ type: "error", title: failureTitle });}};@@ -149,7 +142,10 @@ export function BotComposerModelControl({triggerFit="capped"onOpenChange={setPickerOpen}onInstanceModelChange={(instanceId, model) => {- void changeModel(instanceId, model);+ void saveEngine(+ botEngineForModelPick({ previous: selection, instanceId, model, instanceEntries }),+ t("Could not change the model"),+ );}}/>);diff --git a/apps/web/src/components/roster/BotDetailsPanel.test.tsx b/apps/web/src/components/roster/BotDetailsPanel.test.tsxindex df980fb37..21b1978fc 100644--- a/apps/web/src/components/roster/BotDetailsPanel.test.tsx+++ b/apps/web/src/components/roster/BotDetailsPanel.test.tsx@@ -4,12 +4,7 @@ import { createRef } from "react";import { renderToStaticMarkup } from "react-dom/server";import { describe, expect, it } from "vite-plus/test";-import {- BotDetailsPanel,- BotOverview,- parseBotUsageCapInput,- resolveBotUsageCapForProvider,-} from "./BotDetailsPanel";+import { BotDetailsPanel, BotOverview } from "./BotDetailsPanel";import type { Bot } from "./types";const bot: Bot = {@@ -23,7 +18,6 @@ const bot: Bot = {engine: null,sandbox: null,runtimeMode: "full-access",- usageCap: null,voiceEnabled: false,groupId: null,pinned: false,@@ -50,7 +44,6 @@ describe("BotDetailsPanel", () => {expect(markup).toContain("Sandbox");expect(markup).not.toContain('aria-label="Bot name"');expect(markup).not.toContain('aria-label="Bot description"');- expect(markup).not.toContain("Token hard stop");expect(markup).toContain('aria-hidden="true"');expect(markup).toContain('data-state="closed"');expect(markup).toContain('aria-label="Open Akeru bot sidebar"');@@ -71,29 +64,6 @@ describe("BotDetailsPanel", () => {expect(markup).not.toContain("Unavailable");});- it("sets, clears, and rejects invalid hard stops", () => {- expect(parseBotUsageCapInput("50000")).toEqual({- valid: true,- value: { unit: "tokens", limit: 50_000 },- });- expect(parseBotUsageCapInput(" ")).toEqual({ valid: true, value: null });- expect(parseBotUsageCapInput("0")).toEqual({ valid: false, value: null });- expect(parseBotUsageCapInput("1.5")).toEqual({ valid: false, value: null });- });-- it("clears hard stops for occupancy-only providers", () => {- expect(resolveBotUsageCapForProvider("50000", "grok")).toEqual({- available: false,- valid: true,- value: null,- });- expect(resolveBotUsageCapForProvider("50000", "codex")).toEqual({- available: true,- valid: true,- value: { unit: "tokens", limit: 50_000 },- });- });-it("uses the configured right-panel shortcut while open or closed", () => {const source = NodeFS.readFileSync(new URL("./BotDetailsPanel.tsx", import.meta.url), "utf8");diff --git a/apps/web/src/components/roster/BotDetailsPanel.tsx b/apps/web/src/components/roster/BotDetailsPanel.tsxindex d8abe9c7d..53eeadec2 100644--- a/apps/web/src/components/roster/BotDetailsPanel.tsx+++ b/apps/web/src/components/roster/BotDetailsPanel.tsx@@ -31,11 +31,7 @@ import type { Bot } from "./types";import { useDetailsPanelState } from "./useDetailsPanelState";import { useBotDetailsOpen } from "./detailsPanelOpen";-export {- parseBotUsageCapInput,- resolveBotUsageCapForProvider,- type BotProfileUpdate,-} from "./useBotProfileDraft";+export { type BotProfileUpdate } from "./useBotProfileDraft";export function BotOverview({bot,diff --git a/apps/web/src/components/roster/BotEngineFields.tsx b/apps/web/src/components/roster/BotEngineFields.tsxindex 14b129337..72963b618 100644--- a/apps/web/src/components/roster/BotEngineFields.tsx+++ b/apps/web/src/components/roster/BotEngineFields.tsx@@ -4,13 +4,12 @@ import { useI18n } from "../../i18n";import { ProviderUnavailableNotice } from "../chat/ProviderUnavailableNotice";import { TraitsPicker } from "../chat/TraitsPicker";import { SettingsRow, SettingsSection } from "../settings/settingsLayout";-import { Input } from "../ui/input";import { BotModelPicker } from "./BotModelPicker";import { BotUsageSection } from "./BotUsageSection";import type { Bot } from "./types";import type { BotProfileDraft } from "./useBotProfileDraft";-/** The Model & usage section of bot settings: model, reasoning, token cap, and spend. */+/** The Model & usage section of bot settings: model, reasoning, and spend. */export function BotEngineFields({bot,draft,@@ -71,39 +70,6 @@ export function BotEngineFields({/>) : null}- {draft.resolvedUsageCap.available ? (- <SettingsRow- title={t("Token hard stop")}- description={t(- "Stop this bot once it has spent this many tokens. Leave empty for no limit.",- )}- control={- <Input- className="w-full sm:w-40"- aria-label={t("Token hard stop")}- type="number"- inputMode="numeric"- min={1}- step={1}- value={draft.usageCap}- placeholder={t("No limit")}- onChange={(event) => {- draft.setUsageCap(event.currentTarget.value);- draft.markChanged();- }}- />- }- />- ) : (- <SettingsRow- title={t("Token hard stop")}- description={t(- "This provider reports occupancy rather than tokens, so a token limit does not apply.",- )}- control={<span className="text-sm text-muted-foreground">{t("Unavailable")}</span>}- />- )}-<SettingsRow title={t("Usage")} description={t("What this bot has spent so far.")}><div className="mt-3 max-w-2xl pb-3.5"><BotUsageSection environmentId={environmentId} botId={bot.id} />diff --git a/apps/web/src/components/roster/BotRosterSidebar.tsx b/apps/web/src/components/roster/BotRosterSidebar.tsxindex 87ffcad57..870a9f75f 100644--- a/apps/web/src/components/roster/BotRosterSidebar.tsx+++ b/apps/web/src/components/roster/BotRosterSidebar.tsx@@ -405,7 +405,6 @@ export default function BotRosterSidebar({ chrome = "full" }: { chrome?: "full"engine: null,sandbox: null,runtimeMode: DEFAULT_BOT_RUNTIME_MODE,- usageCap: null,groupId: null,},});diff --git a/apps/web/src/components/roster/BotSettingsPage.imageProvider.test.tsx b/apps/web/src/components/roster/BotSettingsPage.imageProvider.test.tsxindex 37787f4c7..bee336fa4 100644--- a/apps/web/src/components/roster/BotSettingsPage.imageProvider.test.tsx+++ b/apps/web/src/components/roster/BotSettingsPage.imageProvider.test.tsx@@ -148,7 +148,6 @@ function makeBot(overrides: Partial<Bot> = {}): Bot {engine: { provider: codexId, model: "gpt-5" },sandbox: "local",runtimeMode: "full-access",- usageCap: null,personalityTone: 50,voiceEnabled: false,imageProvider: null,@@ -260,7 +259,6 @@ function expectedUpdate(bot: Bot, overrides: TestProps) {label: null,description: null,engine: bot.engine,- usageCap: null,sandbox: bot.sandbox,personalityTone: 50,voiceEnabled: false,diff --git a/apps/web/src/components/roster/BotSettingsPage.zh-CN.test.tsx b/apps/web/src/components/roster/BotSettingsPage.zh-CN.test.tsxindex 700ec31e9..5fa0b4dee 100644--- a/apps/web/src/components/roster/BotSettingsPage.zh-CN.test.tsx+++ b/apps/web/src/components/roster/BotSettingsPage.zh-CN.test.tsx@@ -140,7 +140,6 @@ function makeBot(overrides: Partial<Bot> = {}): Bot {engine: { provider: codexId, model: "gpt-5" },sandbox: "local",runtimeMode: "full-access",- usageCap: null,personalityTone: 50,voiceEnabled: false,imageProvider: null,diff --git a/apps/web/src/components/roster/BotStepMeter.tsx b/apps/web/src/components/roster/BotStepMeter.tsxindex 13776e851..de2ff64d7 100644--- a/apps/web/src/components/roster/BotStepMeter.tsx+++ b/apps/web/src/components/roster/BotStepMeter.tsx@@ -13,7 +13,6 @@ export function BotStepMeter({ meter }: { readonly meter: BotStepMeterData | undformatBotStepEngine(meter.engine),meter.tokens === null ? null : t("{tokens} tokens", { tokens: formatTokens(meter.tokens) }),meter.costUsd === null ? null : formatUsd(meter.costUsd),- meter.hardStopReached ? t("Hard stop") : null,].filter((part) => part !== null);return (diff --git a/apps/web/src/components/roster/BotUsageSection.test.ts b/apps/web/src/components/roster/BotUsageSection.test.tsindex bb287c975..cc5175cca 100644--- a/apps/web/src/components/roster/BotUsageSection.test.ts+++ b/apps/web/src/components/roster/BotUsageSection.test.ts@@ -35,7 +35,6 @@ const snapshot = {reflector: { tokens: 100, unavailableEntries: 0 },},entries: [],- usageCap: { unit: "tokens", limit: 20_000 },estimatedCost: { status: "available", usd: 1.25 },subscriptionPool: { status: "available", used: 34, limit: 100, unit: "percent" },} as const;@@ -98,7 +97,7 @@ describe("formatUsageMeasurement", () => {expect(markup).toContain(value);}- expect(markup).toContain("12,345 / 20,000 tokens");+ expect(markup).not.toContain("Cap</span>");expect(markup).toContain("$1.25");expect(markup).toContain("34% of pool");expect(markup).toContain("Reserved");@@ -127,7 +126,6 @@ describe("formatUsageMeasurement", () => {data: {...snapshot,reservedTokens: 0,- usageCap: null,measurements: {...snapshot.measurements,input: { tokens: 0, unavailableEntries: 1 },@@ -148,7 +146,6 @@ describe("formatUsageMeasurement", () => {'>Subscription pool</span><span class="text-right">Unavailable</span>',);expect(markup).toContain("Some provider usage is unavailable.");- expect(markup).toContain("No cap");expect(markup).not.toContain("Reserved</span>");});});diff --git a/apps/web/src/components/roster/BotUsageSection.tsx b/apps/web/src/components/roster/BotUsageSection.tsxindex 6ee07da80..69bf8be2d 100644--- a/apps/web/src/components/roster/BotUsageSection.tsx+++ b/apps/web/src/components/roster/BotUsageSection.tsx@@ -76,15 +76,6 @@ export function BotUsageSection({<span className="text-right">{formatUsageMeasurement(snapshot.measurements.reflector, t, formatNumber)}</span>- <span className="text-muted-foreground">{t("Cap")}</span>- <span className="text-right">- {snapshot.usageCap- ? t("{consumed} / {limit} tokens", {- consumed: formatNumber(snapshot.consumedTokens),- limit: formatNumber(snapshot.usageCap.limit),- })- : t("No cap")}- </span><span className="text-muted-foreground">{t("Estimated cost")}</span><span className="text-right">{snapshot.estimatedCost.status === "available"diff --git a/apps/web/src/components/roster/BotZeroState.tsx b/apps/web/src/components/roster/BotZeroState.tsxindex d2b9259bf..db226da7b 100644--- a/apps/web/src/components/roster/BotZeroState.tsx+++ b/apps/web/src/components/roster/BotZeroState.tsx@@ -52,7 +52,6 @@ export function BotZeroState() {engine: null,sandbox: null,runtimeMode: DEFAULT_BOT_RUNTIME_MODE,- usageCap: null,groupId: null,},});diff --git a/apps/web/src/components/roster/DelegationCard.test.tsx b/apps/web/src/components/roster/DelegationCard.test.tsxindex e02e9cfc9..992db6e25 100644--- a/apps/web/src/components/roster/DelegationCard.test.tsx+++ b/apps/web/src/components/roster/DelegationCard.test.tsx@@ -112,7 +112,6 @@ const parentBot: Bot = {engine: null,sandbox: "local",runtimeMode: "approval-required",- usageCap: null,voiceEnabled: false,groupId: null,pinned: false,@@ -132,7 +131,6 @@ const childBot: Bot = {engine: null,sandbox: "local",runtimeMode: "approval-required",- usageCap: null,voiceEnabled: false,groupId: null,pinned: false,diff --git a/apps/web/src/components/roster/ThreadLanding.delegations.test.tsx b/apps/web/src/components/roster/ThreadLanding.delegations.test.tsxindex 6f9c5c9b4..d649e6d1f 100644--- a/apps/web/src/components/roster/ThreadLanding.delegations.test.tsx+++ b/apps/web/src/components/roster/ThreadLanding.delegations.test.tsx@@ -262,7 +262,6 @@ const parentBot: Bot = {engine: null,sandbox: "local",runtimeMode: "approval-required",- usageCap: null,voiceEnabled: false,groupId: "group-1",pinned: false,diff --git a/apps/web/src/components/roster/ThreadLanding.hookOrder.test.tsx b/apps/web/src/components/roster/ThreadLanding.hookOrder.test.tsxindex 53b5f44a4..8218459c5 100644--- a/apps/web/src/components/roster/ThreadLanding.hookOrder.test.tsx+++ b/apps/web/src/components/roster/ThreadLanding.hookOrder.test.tsx@@ -276,7 +276,6 @@ const bot: Bot = {engine: null,sandbox: "local",runtimeMode: "approval-required",- usageCap: null,voiceEnabled: false,groupId: null,pinned: false,diff --git a/apps/web/src/components/roster/botEngineSelection.test.ts b/apps/web/src/components/roster/botEngineSelection.test.tsindex 08b713b30..cae1699a6 100644--- a/apps/web/src/components/roster/botEngineSelection.test.ts+++ b/apps/web/src/components/roster/botEngineSelection.test.ts@@ -1,5 +1,10 @@import { catalogRegistry, createTranslator } from "@akeru/client-runtime/i18n";-import { ProviderDriverKind, ProviderInstanceId } from "@akeru/contracts";+import {+ ProviderDriverKind,+ ProviderInstanceId,+ type ModelCapabilities,+ type ServerProvider,+} from "@akeru/contracts";import { DEFAULT_UNIFIED_SETTINGS } from "@akeru/contracts/settings";import { describe, expect, it } from "vite-plus/test";@@ -7,7 +12,10 @@ import { deriveProviderInstanceEntries } from "../../providerInstances";import { makeComposerTestProvider } from "../../test/composerTestProvider";import { providerCatalogEntry } from "../settings/providerCatalog";import {+ botDraftModelOptions,botEngineFailureContext,+ botEngineForModelPick,+ botEngineForOptions,botEngineTakesDelegatedWork,botEngineUnavailability,resolveStickyBotEngine,@@ -223,7 +231,7 @@ describe("resolveStickyBotEngine", () => {expect(resolved?.instanceId).toBe(instanceId);});- it("inherits app options when an older bot engine matches the app model", () => {+ it("keeps a saved engine without options on provider defaults even on the app model", () => {const providers = [makeComposerTestProvider()];const instanceEntries = deriveProviderInstanceEntries(providers);const instanceId = instanceEntries[0]?.instanceId;@@ -242,9 +250,31 @@ describe("resolveStickyBotEngine", () => {options: [{ id: "reasoningEffort", value: "medium" }],},}),- ).toEqual({+ ).toEqual({ instanceId, model: "gpt-5.6-sol" });+ });++ it("lets a bot without an engine show the app default options", () => {+ const providers = [makeComposerTestProvider()];+ const instanceEntries = deriveProviderInstanceEntries(providers);+ const instanceId = instanceEntries[0]?.instanceId;++ if (!instanceId) throw new Error("missing instance");++ const resolved = resolveStickyBotEngine({+ engine: null,+ instanceEntries,+ settings,+ providers,+ defaultSelection: {+ instanceId,+ model: "gpt-5-codex",+ options: [{ id: "reasoningEffort", value: "medium" }],+ },+ });++ expect(resolved).toEqual({instanceId,- model: "gpt-5.6-sol",+ model: "gpt-5-codex",options: [{ id: "reasoningEffort", value: "medium" }],});});@@ -404,3 +434,122 @@ describe("routineDelegateOptions", () => {]);});});++const reasoningCaps = (levels: ReadonlyArray<string>): ModelCapabilities => ({+ optionDescriptors: [+ {+ id: "reasoningEffort",+ label: "Reasoning",+ type: "select",+ currentValue: "default",+ options: [+ { id: "default", label: "Provider default", isDefault: true },+ ...levels.map((id) => ({ id, label: id })),+ ],+ },+ ],+});++function reasoningProvider(instance: string): ServerProvider {+ return {+ ...makeComposerTestProvider(),+ instanceId: ProviderInstanceId.make(instance),+ models: [+ {+ slug: "gpt-a",+ name: "A",+ isCustom: false,+ capabilities: reasoningCaps(["minimal", "high"]),+ },+ { slug: "gpt-b", name: "B", isCustom: false, capabilities: reasoningCaps(["high"]) },+ {+ slug: "gpt-quiet",+ name: "Quiet",+ isCustom: false,+ capabilities: { optionDescriptors: [] },+ },+ { slug: "gpt-custom", name: "Custom", isCustom: true, capabilities: null },+ ],+ };+}++describe("bot engine reasoning choices", () => {+ const instanceEntries = deriveProviderInstanceEntries([+ reasoningProvider("codex"),+ reasoningProvider("codex_work"),+ ]);++ const codex = ProviderInstanceId.make("codex");++ const saved = {+ instanceId: codex,+ model: "gpt-a",+ options: [{ id: "reasoningEffort", value: "minimal" }],+ };++ it("starts the settings draft from the saved engine, not the app default", () => {+ const appDefault = {+ instanceId: codex,+ model: "gpt-a",+ options: [{ id: "reasoningEffort", value: "high" }],+ };++ expect(+ botDraftModelOptions({ provider: "codex", model: "gpt-a" }, appDefault, "codex", "gpt-a"),+ ).toBeUndefined();+ expect(botDraftModelOptions(null, appDefault, "codex", "gpt-a")).toEqual(appDefault.options);+ expect(botDraftModelOptions(null, appDefault, "codex", "gpt-b")).toBeUndefined();+ });++ it("keeps a supported choice when the model changes on the same instance", () => {+ expect(+ botEngineForModelPick({+ previous: saved,+ instanceId: "codex",+ model: "gpt-a",+ instanceEntries,+ }),+ ).toEqual({ provider: "codex", model: "gpt-a", options: saved.options });+ expect(+ botEngineForModelPick({+ previous: saved,+ instanceId: "codex",+ model: "gpt-b",+ instanceEntries,+ }),+ ).toEqual({ provider: "codex", model: "gpt-b" });+ });++ it("clears choices for another instance, an optionless model, and an unknown model", () => {+ for (const [instanceId, model] of [+ ["codex_work", "gpt-a"],+ ["codex", "gpt-quiet"],+ ["codex", "gpt-custom"],+ ] as const) {+ expect(+ botEngineForModelPick({ previous: saved, instanceId, model, instanceEntries }),+ ).toEqual({ provider: instanceId, model });+ }+ });++ it("saves a reasoning change and drops the provider default", () => {+ expect(+ botEngineForOptions({+ selection: saved,+ options: [{ id: "reasoningEffort", value: "high" }],+ instanceEntries,+ }),+ ).toEqual({+ provider: "codex",+ model: "gpt-a",+ options: [{ id: "reasoningEffort", value: "high" }],+ });+ expect(+ botEngineForOptions({+ selection: saved,+ options: [{ id: "reasoningEffort", value: "default" }],+ instanceEntries,+ }),+ ).toEqual({ provider: "codex", model: "gpt-a" });+ });+});diff --git a/apps/web/src/components/roster/botEngineSelection.ts b/apps/web/src/components/roster/botEngineSelection.tsindex f31d4042b..0428b9158 100644--- a/apps/web/src/components/roster/botEngineSelection.ts+++ b/apps/web/src/components/roster/botEngineSelection.ts@@ -3,7 +3,9 @@ import {ProviderDriverKind,ProviderInstanceId,type BotEngine,+ type ModelCapabilities,type ModelSelection,+ type ProviderOptionSelection,type ServerProvider,type ServerProviderUnavailability,type UnifiedSettings,@@ -11,11 +13,16 @@ import {import { createTranslator } from "@akeru/client-runtime/i18n";import { driverSupportsDelegation } from "@akeru/shared/delegationProviders";+import {+ botEngineFromModelSelection,+ botEngineModelSelection,+ providerOptionsForModelChange,+} from "@akeru/shared/model";import type { ProviderAvailabilityTranslate } from "@akeru/client-runtime/provider-availability";import { resolveAppModelSelectionForInstance } from "../../modelSelection";import type { ComposerProviderCatalog } from "../chat/composerProviderMenuItems";-import { formatProviderDriverKindLabel } from "../../providerModels";+import { findProviderModelCapabilities, formatProviderDriverKindLabel } from "../../providerModels";import { providerCatalogEntryForDriver } from "../settings/providerCatalog";import {providerInstanceUnavailability,@@ -70,7 +77,8 @@ export function routineDelegateOptions(* The engine a bot answers with. A saved engine is returned as saved, even when* its provider is signed out, turned off, or no longer lists the model: the* bot keeps its choice and `botEngineUnavailability` explains why it cannot- * run. Only a bot without an engine borrows the app's selectable default.+ * run. Its saved options are its own, even when it runs the app's default model.+ * Only a bot without an engine borrows the app's selectable default.*/export function resolveStickyBotEngine(input: {readonly engine: BotEngine | null;@@ -79,22 +87,7 @@ export function resolveStickyBotEngine(input: {readonly providers: ReadonlyArray<ServerProvider>;readonly defaultSelection: ModelSelection;}): ModelSelection | null {- if (input.engine) {- const instanceId = ProviderInstanceId.make(input.engine.provider);-- const options =- input.engine.options ??- (input.defaultSelection.instanceId === instanceId &&- input.defaultSelection.model === input.engine.model- ? input.defaultSelection.options- : undefined);-- return {- instanceId,- model: input.engine.model,- ...(options ? { options } : {}),- };- }+ if (input.engine) return botEngineModelSelection(input.engine);const entry = resolveSelectableProviderInstanceEntry(input.instanceEntries,@@ -118,6 +111,74 @@ export function resolveStickyBotEngine(input: {};}+/**+ * The options a bot's settings draft starts from: a saved engine's own, or the+ * app default's for a bot without an engine that runs the app default model.+ */+export function botDraftModelOptions(+ engine: BotEngine | null,+ defaultSelection: ModelSelection,+ instanceId: string,+ model: string,+): BotEngine["options"] {+ if (engine) return engine.options;++ return defaultSelection.instanceId === instanceId && defaultSelection.model === model+ ? defaultSelection.options+ : undefined;+}++/** What this client knows about a model's options; undefined when it cannot know. */+function botEngineModelCapabilities(+ instanceEntries: ReadonlyArray<ProviderInstanceEntry>,+ instanceId: string,+ model: string,+): ModelCapabilities | undefined {+ const entry = instanceEntries.find((candidate) => candidate.instanceId === instanceId);++ return entry ? findProviderModelCapabilities(entry.models, model, entry.driverKind) : undefined;+}++/**+ * The engine a model pick saves. The same instance keeps the reasoning choices+ * the new model supports; another instance starts on provider defaults.+ */+export function botEngineForModelPick(input: {+ readonly previous: ModelSelection | null;+ readonly instanceId: string;+ readonly model: string;+ readonly instanceEntries: ReadonlyArray<ProviderInstanceEntry>;+}): BotEngine {+ const options = providerOptionsForModelChange({+ previous: input.previous,+ instanceId: input.instanceId,+ nextCaps: botEngineModelCapabilities(input.instanceEntries, input.instanceId, input.model),+ });++ return { provider: input.instanceId, model: input.model, ...(options ? { options } : {}) };+}++/**+ * The engine a reasoning change saves: the same model with only the choices+ * that differ from its defaults, so picking a default clears that choice.+ */+export function botEngineForOptions(input: {+ readonly selection: ModelSelection;+ readonly options: ReadonlyArray<ProviderOptionSelection> | undefined;+ readonly instanceEntries: ReadonlyArray<ProviderInstanceEntry>;+}): BotEngine {+ const { instanceId, model } = input.selection;++ return botEngineFromModelSelection(+ { instanceId, model, ...(input.options ? { options: input.options } : {}) },+ botEngineModelCapabilities(+ input.instanceEntries,+ input.selection.instanceId,+ input.selection.model,+ ),+ );+}+const englishTranslate: ProviderAvailabilityTranslate = createTranslator("en").t;/**diff --git a/apps/web/src/components/roster/botMessageRowEquality.ts b/apps/web/src/components/roster/botMessageRowEquality.tsindex 6596e5a64..036d4ad95 100644--- a/apps/web/src/components/roster/botMessageRowEquality.ts+++ b/apps/web/src/components/roster/botMessageRowEquality.ts@@ -56,12 +56,7 @@ function stepMetersEqual(a: BotStepMeterData | undefined, b: BotStepMeterData |if (!a || !b) return false;- return (- a.tokens === b.tokens &&- a.costUsd === b.costUsd &&- a.hardStopReached === b.hardStopReached &&- shallowEqual(a.engine, b.engine)- );+ return a.tokens === b.tokens && a.costUsd === b.costUsd && shallowEqual(a.engine, b.engine);}function pluginResultsEqual(diff --git a/apps/web/src/components/roster/roster.test-support.ts b/apps/web/src/components/roster/roster.test-support.tsindex b771568fb..7190ce1d3 100644--- a/apps/web/src/components/roster/roster.test-support.ts+++ b/apps/web/src/components/roster/roster.test-support.ts@@ -10,7 +10,6 @@ export function bot(input: Partial<Bot> & Pick<Bot, "id" | "name">): Bot {engine: null,sandbox: null,runtimeMode: "full-access",- usageCap: null,voiceEnabled: false,groupId: null,pinned: false,diff --git a/apps/web/src/components/roster/roster.zh-CN.test.tsx b/apps/web/src/components/roster/roster.zh-CN.test.tsxindex 0cc2a3c17..59bb01beb 100644--- a/apps/web/src/components/roster/roster.zh-CN.test.tsx+++ b/apps/web/src/components/roster/roster.zh-CN.test.tsx@@ -83,7 +83,6 @@ function makeBot(id: string, name: string): Bot {engine: null,sandbox: null,runtimeMode: "full-access",- usageCap: null,voiceEnabled: false,groupId: null,pinned: false,@@ -141,16 +140,13 @@ describe("roster in Simplified Chinese", () => {engine: { provider: ProviderInstanceId.make("codex"), model: "gpt-5" },tokens: 1200,costUsd: null,- hardStopReached: true,}}/>,);expect(markup).toContain("codex/gpt-5");expect(markup).toContain("个令牌");- expect(markup).toContain("强制停止");expect(markup).not.toContain("tokens");- expect(markup).not.toContain("Hard stop");});it("translates the usage section", () => {diff --git a/apps/web/src/components/roster/rosterRouteSelection.test.ts b/apps/web/src/components/roster/rosterRouteSelection.test.tsindex 1e9858d12..9ea4fb319 100644--- a/apps/web/src/components/roster/rosterRouteSelection.test.ts+++ b/apps/web/src/components/roster/rosterRouteSelection.test.ts@@ -23,7 +23,6 @@ function bot(id: string, archivedAt: string | null = null): Bot {engine: null,sandbox: null,runtimeMode: "full-access",- usageCap: null,voiceEnabled: false,groupId: null,pinned: false,diff --git a/apps/web/src/components/roster/rosterStore.test.ts b/apps/web/src/components/roster/rosterStore.test.tsindex e6b0d03e6..3b1b57caf 100644--- a/apps/web/src/components/roster/rosterStore.test.ts+++ b/apps/web/src/components/roster/rosterStore.test.ts@@ -15,7 +15,6 @@ function bot(id: string, archivedAt: string | null = null): Bot {engine: null,sandbox: null,runtimeMode: "full-access",- usageCap: null,voiceEnabled: false,groupId: null,pinned: false,diff --git a/apps/web/src/components/roster/types.ts b/apps/web/src/components/roster/types.tsindex cf655a58f..7ed2c0af6 100644--- a/apps/web/src/components/roster/types.ts+++ b/apps/web/src/components/roster/types.ts@@ -48,7 +48,6 @@ export interface Bot {| "tenki"| null;runtimeMode: "approval-required" | "auto-accept-edits" | "auto" | "full-access";- usageCap: { unit: "tokens"; limit: number } | null;/** Personality baseline, 0 (chill) to 100 (professional). Absent on bots saved before the field existed. */personalityTone?: number;voiceEnabled: boolean;diff --git a/apps/web/src/components/roster/useBotProfileDraft.ts b/apps/web/src/components/roster/useBotProfileDraft.tsindex 60fc023cf..ff2f74dff 100644--- a/apps/web/src/components/roster/useBotProfileDraft.ts+++ b/apps/web/src/components/roster/useBotProfileDraft.ts@@ -23,7 +23,12 @@ import {} from "../../providerInstances";import { primaryServerProvidersAtom } from "../../state/server";import { shouldRenderTraitsControls } from "../chat/TraitsPicker";-import { botEngineUnavailability } from "./botEngineSelection";+import {+ botDraftModelOptions,+ botEngineForModelPick,+ botEngineForOptions,+ botEngineUnavailability,+} from "./botEngineSelection";import { canonicalizeBotPersonalityTone } from "./botPersonalityTone";import { botSandboxChoice, type BotSandboxChoice } from "./botSandbox";import type { Bot } from "./types";@@ -50,7 +55,6 @@ export interface BotProfileUpdate {readonly label: string | null;readonly description: string | null;readonly engine: Bot["engine"];- readonly usageCap: Bot["usageCap"];readonly sandbox: Bot["sandbox"];readonly personalityTone: number;readonly voiceEnabled: boolean;@@ -66,23 +70,6 @@ export function botImageProviderFromSelectValue(value: string | null): ImageProvreturn value !== null && isImageProviderId(value) ? value : null;}-export function parseBotUsageCapInput(input: string) {- if (input.trim().length === 0) return { valid: true, value: null };- const limit = Number(input);-- if (!Number.isSafeInteger(limit) || limit <= 0) return { valid: false, value: null };-- return { valid: true, value: { unit: "tokens" as const, limit } };-}--export function resolveBotUsageCapForProvider(input: string, providerDriver?: string) {- if (providerDriver === "grok") {- return { available: false, valid: true, value: null };- }-- return { available: true, ...parseBotUsageCapInput(input) };-}-/*** Draft state for every editable bot field, shared by the in-chat bot panel* and the full bot settings page. Both surfaces read dirtiness and build the@@ -99,7 +86,6 @@ export function useBotProfileDraft(const [name, setName] = useState(bot.name);const [label, setLabel] = useState(bot.label ?? "");const [description, setDescription] = useState(bot.description ?? "");- const [usageCap, setUsageCap] = useState(() => bot.usageCap?.limit.toString() ?? "");const [sandbox, setSandbox] = useState<BotSandboxChoice>(() => botSandboxChoice(bot.sandbox));const [personalityTone, setPersonalityTone] = useState(() =>@@ -161,13 +147,9 @@ export function useBotProfileDraft(defaultSelection.model,);- const [modelOptions, setModelOptions] = useState<BotModelOptions>(- () =>- bot.engine?.options ??- (bot.engine?.provider === defaultSelection.instanceId &&- bot.engine.model === defaultSelection.model- ? defaultSelection.options- : undefined),+ // A saved engine shows its own options, never the app default's.+ const [modelOptions, setModelOptions] = useState<BotModelOptions>(() =>+ botDraftModelOptions(bot.engine, defaultSelection, provider, model),);const modelOptionsByInstance = useMemo(@@ -197,13 +179,14 @@ export function useBotProfileDraft(if (bot.engine?.model) setModel(bot.engine.model);setModelOptions(- bot.engine?.options ??- (bot.engine?.provider === defaultSelection.instanceId &&- bot.engine.model === defaultSelection.model- ? defaultSelection.options- : undefined),+ botDraftModelOptions(+ bot.engine,+ defaultSelection,+ bot.engine?.provider ?? defaultSelection.instanceId,+ bot.engine?.model ?? model,+ ),);- }, [bot.engine, defaultSelection, engineChanged]);+ }, [bot.engine, defaultSelection, engineChanged, model]);useEffect(() => {const previous = previousBot.current;@@ -216,13 +199,6 @@ export function useBotProfileDraft(setDescription((current) =>rebaseUneditedValue(current, previous.description ?? "", bot.description ?? ""),);- setUsageCap((current) =>- rebaseUneditedValue(- current,- previous.usageCap?.limit.toString() ?? "",- bot.usageCap?.limit.toString() ?? "",- ),- );setSandbox((current) =>rebaseUneditedValue(current,@@ -276,11 +252,6 @@ export function useBotProfileDraft(allowPromptInjectedEffort: false,});- const resolvedUsageCap = resolveBotUsageCapForProvider(usageCap, activeEntry?.driverKind);-- const usageCapDirty =- !resolvedUsageCap.valid || resolvedUsageCap.value?.limit !== bot.usageCap?.limit;-const toolOverridesDirty =mcpServerIdsKey(disabledMcpServerIds) !== mcpServerIdsKey(bot.disabledMcpServerIds);@@ -291,14 +262,13 @@ export function useBotProfileDraft(normalizedLabel !== bot.label ||normalizedDescription !== bot.description ||engineChanged ||- usageCapDirty ||sandboxDirty ||personalityTone !== savedTone ||voiceEnabled !== bot.voiceEnabled ||imageProvider !== (bot.imageProvider ?? null) ||toolOverridesDirty;- const canSave = Boolean(onSave) && dirty && name.trim().length > 0 && resolvedUsageCap.valid;+ const canSave = Boolean(onSave) && dirty && name.trim().length > 0;return {// Provider context shared by both model pickers.@@ -317,9 +287,6 @@ export function useBotProfileDraft(setLabel,description,setDescription,- usageCap,- setUsageCap,- resolvedUsageCap,sandbox,setSandbox,personalityTone,@@ -335,18 +302,31 @@ export function useBotProfileDraft(/** Applies a model pick, including the option reset the picker implies. */selectModel: (instanceId: string, nextModel: string) => {+ const engine = botEngineForModelPick({+ previous: {+ instanceId: providerInstanceId,+ model,+ ...(modelOptions ? { options: modelOptions } : {}),+ },+ instanceId,+ model: nextModel,+ instanceEntries,+ });+setProvider(instanceId);setModel(nextModel);- setModelOptions(- defaultSelection.instanceId === instanceId && defaultSelection.model === nextModel- ? defaultSelection.options- : undefined,- );+ setModelOptions(engine.options);setEngineChanged(true);markChanged();},selectModelOptions: (nextOptions: BotModelOptions) => {- setModelOptions(nextOptions);+ setModelOptions(+ botEngineForOptions({+ selection: { instanceId: providerInstanceId, model },+ options: nextOptions,+ instanceEntries,+ }).options,+ );setEngineChanged(true);markChanged();},@@ -364,7 +344,6 @@ export function useBotProfileDraft(label: normalizedLabel,description: normalizedDescription,engine: nextEngine,- usageCap: resolvedUsageCap.value,sandbox: sandbox === "default" ? null : sandbox,personalityTone,voiceEnabled,diff --git a/apps/web/src/components/sidebar/SidebarChrome.test.ts b/apps/web/src/components/sidebar/SidebarChrome.test.tsindex c0616c736..14116e4a6 100644--- a/apps/web/src/components/sidebar/SidebarChrome.test.ts+++ b/apps/web/src/components/sidebar/SidebarChrome.test.ts@@ -74,7 +74,6 @@ describe("sidebar footer", () => {engine: null,sandbox: null,runtimeMode: "approval-required",- usageCap: null,imageProvider: null,voiceEnabled: false,channelBindings: [],diff --git a/apps/web/src/components/voice/VoiceCall.test.tsx b/apps/web/src/components/voice/VoiceCall.test.tsxindex 9a245fcf6..f0feb96a9 100644--- a/apps/web/src/components/voice/VoiceCall.test.tsx+++ b/apps/web/src/components/voice/VoiceCall.test.tsx@@ -30,7 +30,6 @@ const bot: Bot = {engine: null,sandbox: null,runtimeMode: "full-access",- usageCap: null,voiceEnabled: false,groupId: null,pinned: false,diff --git a/apps/web/src/components/voice/VoiceCallProvider.test.tsx b/apps/web/src/components/voice/VoiceCallProvider.test.tsxindex 11f1f75fa..44b814323 100644--- a/apps/web/src/components/voice/VoiceCallProvider.test.tsx+++ b/apps/web/src/components/voice/VoiceCallProvider.test.tsx@@ -14,7 +14,6 @@ const mocks = vi.hoisted(() => ({engine: null,sandbox: null,runtimeMode: "full-access",- usageCap: null,voiceEnabled: true,groupId: null,pinned: false,diff --git a/apps/web/src/providerModels.ts b/apps/web/src/providerModels.tsindex 27bc5de19..a90dae45c 100644--- a/apps/web/src/providerModels.ts+++ b/apps/web/src/providerModels.ts@@ -88,12 +88,24 @@ export function getProviderModelCapabilities(model: string | null | undefined,provider: ProviderDriverKind,): ModelCapabilities {- const slug = normalizeModelSlug(model, provider);+ return findProviderModelCapabilities(models, model, provider) ?? EMPTY_CAPABILITIES;+}- const caps =- models.find((candidate) => candidate.slug === slug)?.capabilities ?? EMPTY_CAPABILITIES;+/**+ * The capabilities a listed model advertises, or undefined when this client+ * cannot know them: the model is not listed, or its snapshot carries no+ * descriptor list. Unknown is not the same as an empty list, which says the+ * model takes no options.+ */+export function findProviderModelCapabilities(+ models: ReadonlyArray<ServerProviderModel>,+ model: string | null | undefined,+ provider: ProviderDriverKind,+): ModelCapabilities | undefined {+ const slug = normalizeModelSlug(model, provider);+ const caps = models.find((candidate) => candidate.slug === slug)?.capabilities;- return withoutPlanAgentOption(caps);+ return caps?.optionDescriptors === undefined ? undefined : withoutPlanAgentOption(caps);}// Plan mode is retired, so the opencode "plan" agent is never selectable.diff --git a/docs/internals/providers.md b/docs/internals/providers.mdindex c40d76938..e146dac85 100644--- a/docs/internals/providers.md+++ b/docs/internals/providers.md@@ -202,16 +202,45 @@ The model list comes from the [model catalog](#model-catalog) and `ModelCatalog.Codex combines catalog IDs with bundled and historical compatibility models, and takes names,reasoning efforts, and Fast availability from the catalog entry. Models the catalog does not listfall back to the harness SDK's thinking levels and keep the Standard and Fast tiers. Catalog-efforts the SDK would silently downgrade, such as Max on older GPT aliases, are not offered. The catalog+efforts are sent without the SDK's thinking-level coercion, preserving levels such as Minimal and Max. The catalogclassifies models as current or legacy; it is not an exhaustive allowlist. The OAuth transportforwards the selected ID to the Codex API without a local catalog restriction. Account access is-still checked by the provider when a request runs. Claude merges catalog additions into its-built-in capability catalog without dropping historical models. A Claude model newer than the-build inherits the capabilities of the newest built-in model in its family. Grok includes its API model IDs+still checked by the provider when a request runs. Claude keeps historical models and takes native+effort choices from the catalog, retaining built-in context-window and manual-thinking controls.+When a refresh omits a historical model, its bundled catalog entry or known native built-in levels+remain available.+CLI workflow choices such as Ultracode and Ultrathink are excluded from its native effort picker.+Grok includes its API model IDsalongside catalog additions. It labels the compatibility `grok-build` selection as Grok 4.6 and maps it to `grok-4.6` in`mastraModelId`, because the product slug is not an API model ID. Keeping the selection slug letsexisting bots and the default model pass catalog validation. Custom models are retained.+### Reasoning selections++Model snapshots expose reasoning through `ModelCapabilities.optionDescriptors`. Codex, Claude,+Grok, Kimi For Coding, and OpenCode Go use catalog effort levels; standard OpenCode advertises its+live variant choices after a Provider default entry, because OpenCode resolves its own default from+user config. A model with no adjustable effort publishes an empty descriptor list.+Unknown capabilities remain distinct from an authoritative empty list.++Bot settings on web and the desktop renderer, and the mobile settings sheet, save reasoning to+`bot.engine.options`. Web chat composers do not expose reasoning controls. A saved+engine does not inherit app-default options. Choosing a descriptor's default removes that option;+switching models retains supported choices on the same instance and clears them across instances.+Group chats retain their thread selection and do not edit a responder's engine.++`AgentController.inspectEngine` validates reasoning against a settled snapshot after resolving the+bot override. `ReasoningOptions.nativeModelOptions` maps selections to the Mastra state, and+`withAkeruModelRunOptions` supplies native options on each run: OpenAI Responses effort, Anthropic+effort/thinking, xAI reasoning effort, OpenCode Go's protocol-specific fields, or a Custom API+`reasoning_effort` under the `custom-openai` provider options. Kimi uses adaptive+thinking; older Claude models keep manual thinking where supported. Reset and session reuse replace+the previous state options, so an earlier effort cannot stick. Codex OAuth retains the SDK's Medium+default when no level is selected; API-key requests omit the effort. Standard OpenCode continues+forwarding its selected `variant` through the legacy adapter and omits it at Provider default.+Saved Claude `ultrathink` and `ultracode` choices from the CLI menu pass validation and run at+provider default on the harness path.+Successful credential mutations through poll, completion, and logout RPCs reconcile defaultprovider settings and refresh each affected saved-credential instance before returning. Theregistry publishes the updated snapshot immediately, including for named instances and whendiff --git a/docs/user/bots.md b/docs/user/bots.mdindex 8e4310213..8acaa5e42 100644--- a/docs/user/bots.md+++ b/docs/user/bots.md@@ -36,6 +36,16 @@ next message. Select it to change the bot's model without leaving the chat. Theshows providers that cannot run right now, such as one that is signed out or turned off. Theirmodels are dimmed and say why.+Reasoning controls how much thinking the bot spends before answering. Change it under **Reasoning**+in the bot's settings on web or desktop. On mobile, open **Model and reasoning settings** in the+chat. The choices depend on the selected model: some offer levels, some offer a thinking switch,+and models with fixed reasoning have no control.++The choice is saved with the bot and appears on other clients connected to the same environment.+Select **Provider default** to clear a chosen level and use the provider runtime's default. Changing+models on the same provider keeps choices the new model supports; changing provider starts with+its defaults. Group chat controls do not change a member bot's saved reasoning.+A bot keeps its model when that model stops working. Akeru never switches it for you. Instead, thechat and the bot's settings mark the model unavailable, and **Send** stays off until you fix thecause or pick another model. A line below the composer names the provider, says what went wrong,@@ -48,7 +58,7 @@ and gives one next step. **Set up a provider** opens **Settings > Providers**.| Account is not connected, or its sign-in expired | Connect or reconnect it in **Settings > Providers** || Provider no longer offers the model | Pick another model for the bot || Your provider plan hit its usage or rate limit | Wait for the limit to reset |-| The bot hit its Akeru usage cap | Raise the cap in the bot's settings |+| Your provider account hit its quota or spending cap | Raise it with the provider, or switch providers |When a provider reports that it rerouted a message to a different model while the chat is running,the work log notes it as `Model rerouted from <requested> to <effective>`. Providers that run@@ -74,9 +84,8 @@ Avatars rest on a still frame. A working bot sways, glances down at its work, anits face around its body. Point at a bot and its eyes widen and follow the pointer. Now and then oneresting bot on screen blinks or glances around. With reduced motion turned on, avatars stay still.-Set **Token hard stop** to interrupt the current step when it reaches the selected limit. A settled-reply shows its engine, step tokens, and estimated USD cost when the provider reports enough usage-data.+A settled reply shows its engine, step tokens, and estimated USD cost when the provider reports+enough usage data.Akeru stores bot profiles on the connected environment. Every client connected to that environmentsees the same profile.@@ -204,8 +213,8 @@ work that is still running always stay. For finished, failed, or canceled work,read the full result.When a request fails, the chat shows a short card that says what went wrong and what to do next. A-provider that is turned off or signed out gets a **Settings > Providers** link. An Akeru usage cap-gets a **Bot settings** button. When the cause is unknown, select **Send feedback** to report it+provider that is turned off or signed out gets a **Settings > Providers** link. A provider account+limit gets a **Bot settings** button so you can switch the bot to another provider. When the cause is unknown, select **Send feedback** to report itwith its details. Expand **Technical details** to see the underlying error, or select **Resume**when an interrupted request can continue. If your message was saved but the bot never started, thechat says so under your message instead of leaving the reply empty. Group chats do the same and name the bot that was asked to reply. On mobile, the resume carddiff --git a/docs/user/usage.md b/docs/user/usage.mdindex 1f96ac26d..c0e5909db 100644--- a/docs/user/usage.md+++ b/docs/user/usage.md@@ -42,8 +42,8 @@ Web and desktop charts stay still when idle. Hover over a chart to inspect a val## Per-bot usageOn mobile, **Settings > General > Usage** lists the bots in each connected environment below the plan-limits. Open a bot to see its input, output, Observer, and Reflector tokens, its cap, an estimated-cost, subscription pool use, and reserved tokens. Archived bots remain in the list with an+limits. Open a bot to see its input, output, Observer, and Reflector tokens, an estimated cost,+subscription pool use, and reserved tokens. Archived bots remain in the list with an**archived** label so you can review their earlier usage.A measurement a provider did not report reads **Unavailable** rather than zero, and a measurement@@ -54,8 +54,6 @@ Estimated cost comes from the model rate table. It is not subscription spend andbilled. Subscription pool use is provider-reported as either a percentage or a token count; when aprovider reports no meter, the row reads Unavailable.-The cap is read-only here. Change a bot's cap in a chat with that bot, under chat settings.-Per-bot usage reads when you open it, when the app returns to the foreground, and when you pull torefresh. It does not poll in the background.diff --git a/packages/client-runtime/src/bot-step-usage.test.ts b/packages/client-runtime/src/bot-step-usage.test.tsindex 6ed8e4b2e..586633c2d 100644--- a/packages/client-runtime/src/bot-step-usage.test.ts+++ b/packages/client-runtime/src/bot-step-usage.test.ts@@ -20,7 +20,7 @@ function activity(}describe("bot step meter", () => {- it("formats settled step usage and a hard stop", () => {+ it("formats settled step usage", () => {const meters = buildBotStepMeters([activity("usage", "bot.step-usage.updated", {botId: "bot-1",@@ -28,14 +28,12 @@ describe("bot step meter", () => {tokens: 1_200,estimatedCost: { status: "available", usd: 0.42 },}),- activity("cap", "bot.usage-cap.hit", { limit: 1_000 }),]);expect(meters.get("turn-1")).toEqual({engine: { provider: "codex", model: "gpt-5.6-sol" },tokens: 1_200,costUsd: 0.42,- hardStopReached: true,});});@@ -52,7 +50,7 @@ describe("bot step meter", () => {expect(meters.get("turn-1")).toMatchObject({ tokens: null, costUsd: null });expect(formatBotStepEngine(meters.get("turn-1")!.engine)).toBe("anthropic/claude-opus-5");});- it("applies a cap hit before updates and keeps the last valid snapshot", () => {+ it("keeps the last valid snapshot", () => {const snapshot = {botId: "bot-1",engine: { provider: "codex", model: "gpt-5.6-sol" },@@ -61,7 +59,6 @@ describe("bot step meter", () => {};const meters = buildBotStepMeters([- activity("cap", "bot.usage-cap.hit", {}),activity("usage1", "bot.step-usage.updated", snapshot),activity("usage2", "bot.step-usage.updated", { ...snapshot, tokens: 200 }),activity("invalid", "bot.step-usage.updated", { tokens: "invalid" }),@@ -69,8 +66,7 @@ describe("bot step meter", () => {]);expect(meters.size).toBe(1);- expect(meters.get("turn-1")).toMatchObject({ tokens: 200, hardStopReached: true });+ expect(meters.get("turn-1")).toMatchObject({ tokens: 200 });expect(formatBotStepEngine(snapshot.engine)).toBe("codex/gpt-5.6-sol");- expect(buildBotStepMeters([activity("cap", "bot.usage-cap.hit", {})]).size).toBe(0);});});diff --git a/packages/client-runtime/src/bot-step-usage.ts b/packages/client-runtime/src/bot-step-usage.tsindex a34d41ed9..ba9b62ce2 100644--- a/packages/client-runtime/src/bot-step-usage.ts+++ b/packages/client-runtime/src/bot-step-usage.ts@@ -11,21 +11,14 @@ export interface BotStepMeterData {readonly engine: BotEngine;readonly tokens: number | null;readonly costUsd: number | null;- readonly hardStopReached: boolean;}export function buildBotStepMeters(activities: ReadonlyArray<OrchestrationThreadActivity>,): ReadonlyMap<string, BotStepMeterData> {const meters = new Map<string, BotStepMeterData>();- const cappedTurns = new Set<string>();for (const activity of activities) {- if (activity.kind === "bot.usage-cap.hit" && activity.turnId !== null) {- cappedTurns.add(activity.turnId);- continue;- }-if (activity.kind !== "bot.step-usage.updated" ||activity.turnId === null ||@@ -41,16 +34,9 @@ export function buildBotStepMeters(activity.payload.estimatedCost.status === "available"? activity.payload.estimatedCost.usd: null,- hardStopReached: false,});}- for (const turnId of cappedTurns) {- const meter = meters.get(turnId);-- if (meter) meters.set(turnId, { ...meter, hardStopReached: true });- }-return meters;}diff --git a/packages/client-runtime/src/i18n/__snapshots__/coverage.test.ts.snap b/packages/client-runtime/src/i18n/__snapshots__/coverage.test.ts.snapindex dc35034ba..1368319a1 100644--- a/packages/client-runtime/src/i18n/__snapshots__/coverage.test.ts.snap+++ b/packages/client-runtime/src/i18n/__snapshots__/coverage.test.ts.snap@@ -2,7 +2,7 @@exports[`discovered interface message coverage > covers static calls and finite label sources without translating identifiers 1`] = `{- "directMessageCount": 1776,+ "directMessageCount": 1766,"expressions": {"apps/mobile/src/components/DictationControls.tsx": ["announcements[status]);",@@ -363,6 +363,7 @@ exports[`discovered interface message coverage > covers static calls and finite"apps/mobile/src/features/threads/ThreadNavigationSidebar.tsx","apps/mobile/src/features/threads/ThreadRouteScreen.tsx","apps/mobile/src/features/threads/ThreadRoutinesScreen.tsx",+ "apps/mobile/src/features/threads/ThreadSettingsSheet.tsx","apps/mobile/src/features/threads/new-task-flow-provider.tsx","apps/mobile/src/features/threads/sidebar-header-actions.tsx","apps/mobile/src/features/threads/thread-detail-status.tsx",@@ -378,6 +379,7 @@ exports[`discovered interface message coverage > covers static calls and finite"apps/mobile/src/features/threads/thread-settings-picker.tsx","apps/mobile/src/features/threads/thread-work-log.tsx","apps/mobile/src/features/threads/use-new-task-share-import.ts",+ "apps/mobile/src/features/threads/use-thread-composer-settings.ts","apps/mobile/src/features/updates/app-updates.ts","apps/mobile/src/lib/useEnvironmentComposerDictation.ts","apps/mobile/src/settings-stack.tsx",diff --git a/packages/client-runtime/src/i18n/catalogs/en/common-labels.ts b/packages/client-runtime/src/i18n/catalogs/en/common-labels.tsindex 413daf735..849a73863 100644--- a/packages/client-runtime/src/i18n/catalogs/en/common-labels.ts+++ b/packages/client-runtime/src/i18n/catalogs/en/common-labels.ts@@ -45,7 +45,6 @@ export const commonLabelsCatalog = {"Next action": "Next action","Next question": "Next question","No cached data": "No cached data",- "No limit": "No limit","No results": "No results","Open new terminal": "Open new terminal","Open sign-in": "Open sign-in",@@ -229,7 +228,6 @@ export const commonLabelsCatalog = {"Dismiss error": "Dismiss error","Entering text": "Entering text","Failure details unavailable": "Failure details unavailable",- "Hard stop": "Hard stop",Install: "Install","Loading errors": "Loading errors",Manage: "Manage",@@ -286,8 +284,6 @@ export const commonLabelsCatalog = {Input: "Input",Output: "Output",Reflector: "Reflector",- Cap: "Cap",- "No cap": "No cap","Estimated cost": "Estimated cost",Reserved: "Reserved",Chill: "Chill",@@ -409,7 +405,6 @@ export const commonLabelsCatalog = {"Ask first": "Ask first","View diagnostics": "View diagnostics","Terminal logs only": "Terminal logs only",- "Akeru usage cap reached": "Akeru usage cap reached","Unlocks with": "Unlocks with","API access": "API access","Published limits": "Published limits",diff --git a/packages/client-runtime/src/i18n/catalogs/en/connections.ts b/packages/client-runtime/src/i18n/catalogs/en/connections.tsindex c8e7cf999..8907b3bb6 100644--- a/packages/client-runtime/src/i18n/catalogs/en/connections.ts+++ b/packages/client-runtime/src/i18n/catalogs/en/connections.ts@@ -166,8 +166,6 @@ export const connectionsCatalog = {"MCP servers: {count}": "MCP servers: {count}","no sandbox": "no sandbox","{sandbox} sandbox": "{sandbox} sandbox",- "This provider reports occupancy rather than tokens, so a token limit does not apply.":- "This provider reports occupancy rather than tokens, so a token limit does not apply.","Sandbox provider": "Sandbox provider","Which workspace tools this bot may reach.": "Which workspace tools this bot may reach.","No workspace tools": "No workspace tools",diff --git a/packages/client-runtime/src/i18n/catalogs/en/settings.ts b/packages/client-runtime/src/i18n/catalogs/en/settings.tsindex 7daa10de8..aa2db8160 100644--- a/packages/client-runtime/src/i18n/catalogs/en/settings.ts+++ b/packages/client-runtime/src/i18n/catalogs/en/settings.ts@@ -42,8 +42,6 @@ export const settingsCatalog = {"Stream token by token (legacy)": "Stream token by token (legacy)","Terminal font": "Terminal font",Themes: "Themes",- "Token hard stop": "Token hard stop",- "Unavailable for this provider": "Unavailable for this provider","Update ready": "Update ready","Voice provider": "Voice provider","OpenAI API voice": "OpenAI API voice",@@ -80,6 +78,7 @@ export const settingsCatalog = {"No response from {provider}": "No response from {provider}","No response from {provider} for {duration}": "No response from {provider} for {duration}","Could not change the model": "Could not change the model",+ "Could not change the reasoning": "Could not change the reasoning","Change model": "Change model","Nothing stashed yet. Press {shortcut} with a prompt in the composer to stash it.":"Nothing stashed yet. Press {shortcut} with a prompt in the composer to stash it.",@@ -124,8 +123,6 @@ export const settingsCatalog = {"The provider and model this bot runs on.": "The provider and model this bot runs on.","No model yet": "No model yet",Reasoning: "Reasoning",- "Stop this bot once it has spent this many tokens. Leave empty for no limit.":- "Stop this bot once it has spent this many tokens. Leave empty for no limit.","Which subscription this bot uses to create images. The chat model above stays the same.":"Which subscription this bot uses to create images. The chat model above stays the same.","Image provider": "Image provider",@@ -229,8 +226,12 @@ export const settingsCatalog = {"Your {provider} plan hit its usage or rate limit. Wait for it to reset, then send your message again.","Your provider plan hit its usage or rate limit. Wait for it to reset, then send your message again.":"Your provider plan hit its usage or rate limit. Wait for it to reset, then send your message again.",- "Raise this bot's usage cap in its settings to keep chatting.":- "Raise this bot's usage cap in its settings to keep chatting.",+ "{provider} account limit reached": "{provider} account limit reached",+ "Provider account limit reached": "Provider account limit reached",+ "Your {provider} account hit its quota or spending limit. Raise it with {provider}, or switch this bot to another provider in bot settings.":+ "Your {provider} account hit its quota or spending limit. Raise it with {provider}, or switch this bot to another provider in bot settings.",+ "Your provider account hit its quota or spending limit. Raise it with the provider, or switch this bot to another provider in bot settings.":+ "Your provider account hit its quota or spending limit. Raise it with the provider, or switch this bot to another provider in bot settings.","{provider} could not respond": "{provider} could not respond","The provider could not respond": "The provider could not respond",Models: "Models",diff --git a/packages/client-runtime/src/i18n/catalogs/zh-CN/common-labels.ts b/packages/client-runtime/src/i18n/catalogs/zh-CN/common-labels.tsindex 6294a17ef..76d341c60 100644--- a/packages/client-runtime/src/i18n/catalogs/zh-CN/common-labels.ts+++ b/packages/client-runtime/src/i18n/catalogs/zh-CN/common-labels.ts@@ -45,7 +45,6 @@ export const commonLabelsCatalog = {"Next action": "下一步操作","Next question": "下一个问题","No cached data": "没有缓存数据",- "No limit": "无限制","No results": "没有结果","Open new terminal": "打开新终端","Open sign-in": "打开登录",@@ -229,7 +228,6 @@ export const commonLabelsCatalog = {"Dismiss error": "关闭错误","Entering text": "正在输入文本","Failure details unavailable": "失败详情不可用",- "Hard stop": "强制停止",Install: "安装","Loading errors": "正在加载错误",Manage: "管理",@@ -286,8 +284,6 @@ export const commonLabelsCatalog = {Input: "输入",Output: "输出",Reflector: "反思器",- Cap: "上限",- "No cap": "无上限","Estimated cost": "预估费用",Reserved: "已预留",Chill: "随性",@@ -409,7 +405,6 @@ export const commonLabelsCatalog = {"Ask first": "先询问","View diagnostics": "查看诊断信息","Terminal logs only": "仅终端日志",- "Akeru usage cap reached": "已达到 Akeru 用量上限","Unlocks with": "解锁方式","API access": "API 访问","Published limits": "公布的限制",diff --git a/packages/client-runtime/src/i18n/catalogs/zh-CN/connections.ts b/packages/client-runtime/src/i18n/catalogs/zh-CN/connections.tsindex 63e9037bd..8105eef52 100644--- a/packages/client-runtime/src/i18n/catalogs/zh-CN/connections.ts+++ b/packages/client-runtime/src/i18n/catalogs/zh-CN/connections.ts@@ -153,8 +153,6 @@ export const connectionsCatalog = {"MCP servers: {count}": "MCP 服务器:{count}","no sandbox": "无沙箱","{sandbox} sandbox": "{sandbox} 沙箱",- "This provider reports occupancy rather than tokens, so a token limit does not apply.":- "此提供商报告的是占用率而不是令牌,因此令牌上限不适用。","Sandbox provider": "沙箱提供商","Which workspace tools this bot may reach.": "此机器人可以使用哪些工作区工具。","No workspace tools": "没有工作区工具",diff --git a/packages/client-runtime/src/i18n/catalogs/zh-CN/settings.ts b/packages/client-runtime/src/i18n/catalogs/zh-CN/settings.tsindex ea7b94010..5a5f42af6 100644--- a/packages/client-runtime/src/i18n/catalogs/zh-CN/settings.ts+++ b/packages/client-runtime/src/i18n/catalogs/zh-CN/settings.ts@@ -42,8 +42,6 @@ export const settingsCatalog = {"Stream token by token (legacy)": "逐个令牌流式传输(旧版)","Terminal font": "终端字体",Themes: "主题",- "Token hard stop": "令牌硬上限",- "Unavailable for this provider": "此提供商不可用","Update ready": "更新已就绪","Voice provider": "语音提供商","OpenAI API voice": "OpenAI API 语音",@@ -78,6 +76,7 @@ export const settingsCatalog = {"No response from {provider}": "{provider} 暂无响应","No response from {provider} for {duration}": "{provider} 已 {duration} 无响应","Could not change the model": "无法更改模型",+ "Could not change the reasoning": "无法更改推理设置","Change model": "更改模型","Nothing stashed yet. Press {shortcut} with a prompt in the composer to stash it.":"暂无暂存内容。在输入框中输入提示后按 {shortcut} 即可暂存。",@@ -120,8 +119,6 @@ export const settingsCatalog = {"The provider and model this bot runs on.": "此机器人使用的提供商和模型。","No model yet": "尚未选择模型",Reasoning: "推理",- "Stop this bot once it has spent this many tokens. Leave empty for no limit.":- "此机器人用完这么多令牌后即停止。留空表示不限制。","Which subscription this bot uses to create images. The chat model above stays the same.":"此机器人用哪个订阅生成图片。上方的聊天模型保持不变。","Image provider": "图片提供商",@@ -224,8 +221,12 @@ export const settingsCatalog = {"你的 {provider} 套餐已达到用量或速率限制。请等待限额重置后重新发送消息。","Your provider plan hit its usage or rate limit. Wait for it to reset, then send your message again.":"你的提供商套餐已达到用量或速率限制。请等待重置后重新发送消息。",- "Raise this bot's usage cap in its settings to keep chatting.":- "请在此机器人的设置中提高用量上限以继续聊天。",+ "{provider} account limit reached": "{provider} 账户已达到限额",+ "Provider account limit reached": "提供商账户已达到限额",+ "Your {provider} account hit its quota or spending limit. Raise it with {provider}, or switch this bot to another provider in bot settings.":+ "你的 {provider} 账户已达到配额或消费上限。请在 {provider} 提高上限,或在机器人设置中将此机器人切换到其他提供商。",+ "Your provider account hit its quota or spending limit. Raise it with the provider, or switch this bot to another provider in bot settings.":+ "你的提供商账户已达到配额或消费上限。请在提供商处提高上限,或在机器人设置中将此机器人切换到其他提供商。","{provider} could not respond": "{provider} 无法响应","The provider could not respond": "提供商无法响应",Models: "模型",diff --git a/packages/client-runtime/src/providerAvailability.test.ts b/packages/client-runtime/src/providerAvailability.test.tsindex 8989cc1f9..72c1eba38 100644--- a/packages/client-runtime/src/providerAvailability.test.ts+++ b/packages/client-runtime/src/providerAvailability.test.ts@@ -147,8 +147,13 @@ describe("presentProviderUnavailability", () => {action: "providers",});expect(- presentProviderUnavailability({ reason: "usage-cap", providerName: "Codex" }).action,- ).toBe("usage");+ presentProviderUnavailability({ reason: "usage-cap", providerName: "Codex" }),+ ).toMatchObject({+ title: "Codex account limit reached",+ description:+ "Your Codex account hit its quota or spending limit. Raise it with Codex, or switch this bot to another provider in bot settings.",+ action: "usage",+ });});it("renders every reason in the active interface language", () => {diff --git a/packages/client-runtime/src/providerAvailability.ts b/packages/client-runtime/src/providerAvailability.tsindex 0d1ef6c37..17f352e66 100644--- a/packages/client-runtime/src/providerAvailability.ts+++ b/packages/client-runtime/src/providerAvailability.ts@@ -19,7 +19,7 @@ export type ProviderAvailabilityReason =| "not-installed"| "missing-provider";-/** One next step. Clients map "providers" to Settings > Providers and "usage" to bot usage settings. */+/** One next step. Clients map "providers" to Settings > Providers and "usage" to bot settings. */export type ProviderAvailabilityAction = "providers" | "usage" | "feedback" | "none";export interface ProviderAvailabilityPresentation {@@ -185,8 +185,11 @@ function presentReason(};case "usage-cap":return {- title: t("Akeru usage cap reached"),- description: t("Raise this bot's usage cap in its settings to keep chatting."),+ title: named("{provider} account limit reached", "Provider account limit reached"),+ description: named(+ "Your {provider} account hit its quota or spending limit. Raise it with {provider}, or switch this bot to another provider in bot settings.",+ "Your provider account hit its quota or spending limit. Raise it with the provider, or switch this bot to another provider in bot settings.",+ ),technicalDetails,action: "usage",};diff --git a/packages/client-runtime/src/state/botUsage.test.ts b/packages/client-runtime/src/state/botUsage.test.tsindex 8110273e8..d4e1df0ba 100644--- a/packages/client-runtime/src/state/botUsage.test.ts+++ b/packages/client-runtime/src/state/botUsage.test.ts@@ -41,7 +41,6 @@ const snapshotFor = (botId: BotId) => ({reflector: { tokens: 0, unavailableEntries: 0 },},entries: [],- usageCap: null,estimatedCost: { status: "unavailable" as const, usd: null },subscriptionPool: { status: "unavailable" as const, used: null, limit: null, unit: null },});@@ -140,7 +139,6 @@ describe("bot usage environment atoms", () => {reflector: { tokens: 0, unavailableEntries: 0 },},entries: [],- usageCap: null,estimatedCost: { status: "unavailable" as const, usd: null },subscriptionPool: {status: "unavailable" as const,diff --git a/packages/client-runtime/src/state/shellReducer.test-support.ts b/packages/client-runtime/src/state/shellReducer.test-support.tsindex 06796da1d..4df6082b9 100644--- a/packages/client-runtime/src/state/shellReducer.test-support.ts+++ b/packages/client-runtime/src/state/shellReducer.test-support.ts@@ -109,7 +109,6 @@ export const stubBot = {engine: null,sandbox: "local" as const,runtimeMode: "full-access" as const,- usageCap: null,imageProvider: null,voiceEnabled: false,channelBindings: [],diff --git a/packages/contracts/src/akeruUsage.test.ts b/packages/contracts/src/akeruUsage.test.tsindex be31d035a..65022bba5 100644--- a/packages/contracts/src/akeruUsage.test.ts+++ b/packages/contracts/src/akeruUsage.test.ts@@ -21,12 +21,11 @@ it.effect("decodes bot usage with unavailable provider totals", () =>reflector: { tokens: 0, unavailableEntries: 1 },},entries: [],- usageCap: { unit: "tokens", limit: 1_000 },estimatedCost: { status: "unavailable", usd: null },subscriptionPool: { status: "unavailable", used: null, limit: null, unit: null },});- assert.strictEqual(snapshot.usageCap?.limit, 1_000);+ assert.strictEqual(snapshot.consumedTokens, 12);assert.strictEqual(snapshot.estimatedCost.status, "unavailable");}),);diff --git a/packages/contracts/src/akeruUsage.ts b/packages/contracts/src/akeruUsage.tsindex 160a87c72..2b1004823 100644--- a/packages/contracts/src/akeruUsage.ts+++ b/packages/contracts/src/akeruUsage.ts@@ -10,7 +10,7 @@ import {TurnId,} from "./baseSchemas.ts";import { ProviderDriverKind } from "./providerInstance.ts";-import { BotEngine, BotUsageCap } from "./orchestration/roster.ts";+import { BotEngine } from "./orchestration/roster.ts";export const AkeruUsageCategory = Schema.Literals(["turn",@@ -92,7 +92,6 @@ export type AkeruStepUsageSnapshot = typeof AkeruStepUsageSnapshot.Type;export const AkeruBotUsageSnapshot = Schema.Struct({...AkeruBotUsageSummary.fields,- usageCap: Schema.NullOr(BotUsageCap),estimatedCost: AkeruEstimatedCost,subscriptionPool: Schema.Union([Schema.Struct({diff --git a/packages/contracts/src/botConfig.test.ts b/packages/contracts/src/botConfig.test.tsindex aae8dbe85..d673778ef 100644--- a/packages/contracts/src/botConfig.test.ts+++ b/packages/contracts/src/botConfig.test.ts@@ -1,27 +1,14 @@import * as Schema from "effect/Schema";import { describe, expect, it } from "vite-plus/test";-import { BotCreatedPayload, BotPersonalityTone, BotUsageCap } from "./orchestration.ts";--const decodeUsageCap = Schema.decodeUnknownSync(BotUsageCap);+import { BotCreatedPayload, BotPersonalityTone } from "./orchestration.ts";const decodeCreated = Schema.decodeUnknownSync(BotCreatedPayload);const decodePersonalityTone = Schema.decodeUnknownSync(BotPersonalityTone);-describe("BotUsageCap", () => {- it("accepts positive token limits and rejects invalid limits", () => {- expect(decodeUsageCap({ unit: "tokens", limit: 50_000 })).toEqual({- unit: "tokens",- limit: 50_000,- });-- for (const limit of [0, -1, 1.5, Number.NaN, Number.POSITIVE_INFINITY]) {- expect(() => decodeUsageCap({ unit: "tokens", limit })).toThrow();- }- });-- it("defaults old bot events to full access with no cap", () => {+describe("BotCreatedPayload", () => {+ it("defaults old bot events to full access", () => {const bot = decodeCreated({botId: "bot-old",name: "Old bot",@@ -35,7 +22,6 @@ describe("BotUsageCap", () => {});expect(bot.runtimeMode).toBe("full-access");- expect(bot.usageCap).toBeNull();expect(bot.label).toBeNull();expect(bot.description).toBeNull();expect(bot.disabledMcpServerIds).toEqual([]);@@ -43,6 +29,23 @@ describe("BotUsageCap", () => {expect(bot.personalityTone).toBe(50);});+ it("ignores the retired token cap on older bot events", () => {+ const bot = decodeCreated({+ botId: "bot-capped",+ name: "Capped bot",+ title: "Generalist",+ avatar: { kind: "blob", shape: "circle", color: "#5B7FD4" },+ engine: null,+ sandbox: null,+ usageCap: { unit: "tokens", limit: 50_000 },+ groupId: null,+ createdAt: "2026-01-01T00:00:00.000Z",+ updatedAt: "2026-01-01T00:00:00.000Z",+ });++ expect(bot).not.toHaveProperty("usageCap");+ });+it("accepts cloud sandbox providers on bot events", () => {for (const sandbox of ["local",diff --git a/packages/contracts/src/orchestration.roster.test.ts b/packages/contracts/src/orchestration.roster.test.tsindex 9989c759c..2f7c3c194 100644--- a/packages/contracts/src/orchestration.roster.test.ts+++ b/packages/contracts/src/orchestration.roster.test.ts@@ -311,7 +311,6 @@ it.effect("leaves an omitted bot.create runtime mode for the server", () =>avatar: { kind: "dither", seed: "akeru" },engine: null,sandbox: null,- usageCap: null,groupId: null,createdAt: "2026-01-01T00:00:00.000Z",});diff --git a/packages/contracts/src/orchestration.ts b/packages/contracts/src/orchestration.tsindex 096bc9887..e839c9bd0 100644--- a/packages/contracts/src/orchestration.ts+++ b/packages/contracts/src/orchestration.ts@@ -48,7 +48,6 @@ export {PersistedBotSandbox,BotSandboxBrowserSharing,DEFAULT_BOT_SANDBOX_BROWSER_SHARING,- BotUsageCap,MIN_BOT_PERSONALITY_TONE,BALANCED_BOT_PERSONALITY_TONE,MAX_BOT_PERSONALITY_TONE,diff --git a/packages/contracts/src/orchestration/eventPayloads.ts b/packages/contracts/src/orchestration/eventPayloads.tsindex 4b34a8e01..7177d501d 100644--- a/packages/contracts/src/orchestration/eventPayloads.ts+++ b/packages/contracts/src/orchestration/eventPayloads.ts@@ -35,7 +35,6 @@ import {BotAvatar,BotEngine,PersistedBotSandbox,- BotUsageCap,BALANCED_BOT_PERSONALITY_TONE,BotPersonalityTone,ChannelBinding,@@ -189,7 +188,6 @@ export const BotCreatedPayload = Schema.Struct({engine: Schema.NullOr(BotEngine),sandbox: PersistedBotSandbox,runtimeMode: RuntimeMode.pipe(Schema.withDecodingDefault(Effect.succeed(DEFAULT_RUNTIME_MODE))),- usageCap: Schema.NullOr(BotUsageCap).pipe(Schema.withDecodingDefault(Effect.succeed(null))),imageProvider: Schema.NullOr(ImageProviderId).pipe(Schema.withDecodingDefault(Effect.succeed(null)),),@@ -216,7 +214,6 @@ export const BotUpdatedPayload = Schema.Struct({engine: Schema.optional(Schema.NullOr(BotEngine)),sandbox: Schema.optional(PersistedBotSandbox),runtimeMode: Schema.optional(RuntimeMode),- usageCap: Schema.optional(Schema.NullOr(BotUsageCap)),imageProvider: Schema.optional(Schema.NullOr(ImageProviderId)),personalityTone: Schema.optional(BotPersonalityTone),voiceEnabled: Schema.optional(Schema.Boolean),diff --git a/packages/contracts/src/orchestration/roster.ts b/packages/contracts/src/orchestration/roster.tsindex 7554cbb47..185d0b79b 100644--- a/packages/contracts/src/orchestration/roster.ts+++ b/packages/contracts/src/orchestration/roster.ts@@ -143,13 +143,6 @@ export type BotSandboxBrowserSharing = typeof BotSandboxBrowserSharing.Type;export const DEFAULT_BOT_SANDBOX_BROWSER_SHARING: BotSandboxBrowserSharing = "separate";-export const BotUsageCap = Schema.Struct({- unit: Schema.Literal("tokens"),- limit: Schema.Int.check(Schema.isGreaterThan(0)),-});--export type BotUsageCap = typeof BotUsageCap.Type;-export const MIN_BOT_PERSONALITY_TONE = 0;export const BALANCED_BOT_PERSONALITY_TONE = 50;@@ -294,7 +287,6 @@ export const OrchestrationBot = Schema.Struct({engine: Schema.NullOr(BotEngine),sandbox: PersistedBotSandbox,runtimeMode: RuntimeMode.pipe(Schema.withDecodingDefault(Effect.succeed(DEFAULT_RUNTIME_MODE))),- usageCap: Schema.NullOr(BotUsageCap).pipe(Schema.withDecodingDefault(Effect.succeed(null))),/*** The bot's image provider, independent of its chat engine (a Claude bot may* still use ChatGPT images). `null` means "use the global default" and isdiff --git a/packages/contracts/src/orchestration/rosterCommands.ts b/packages/contracts/src/orchestration/rosterCommands.tsindex b66434c3d..90265f1f4 100644--- a/packages/contracts/src/orchestration/rosterCommands.ts+++ b/packages/contracts/src/orchestration/rosterCommands.ts@@ -1,4 +1,3 @@-import * as Effect from "effect/Effect";import * as Schema from "effect/Schema";import { ThreadEnvMode } from "../environment.ts";import {@@ -22,7 +21,6 @@ import {BotAvatar,BotEngine,BotSandbox,- BotUsageCap,BotPersonalityTone,ChannelProvider,ChannelBinding,@@ -74,7 +72,6 @@ export const BotCreateCommand = Schema.Struct({engine: Schema.NullOr(BotEngine),sandbox: Schema.NullOr(BotSandbox),runtimeMode: Schema.optional(RuntimeMode),- usageCap: Schema.NullOr(BotUsageCap).pipe(Schema.withDecodingDefault(Effect.succeed(null))),imageProvider: Schema.optional(Schema.NullOr(ImageProviderId)),personalityTone: Schema.optional(BotPersonalityTone),voiceEnabled: Schema.optional(Schema.Boolean),@@ -95,7 +92,6 @@ export const BotUpdateCommand = Schema.Struct({engine: Schema.optional(Schema.NullOr(BotEngine)),sandbox: Schema.optional(Schema.NullOr(BotSandbox)),runtimeMode: Schema.optional(RuntimeMode),- usageCap: Schema.optional(Schema.NullOr(BotUsageCap)),imageProvider: Schema.optional(Schema.NullOr(ImageProviderId)),personalityTone: Schema.optional(BotPersonalityTone),voiceEnabled: Schema.optional(Schema.Boolean),@@ -116,7 +112,6 @@ export const ClientBotUpdateCommand = Schema.Struct({engine: Schema.optional(Schema.NullOr(BotEngine)),sandbox: Schema.optional(Schema.NullOr(BotSandbox)),runtimeMode: Schema.optional(RuntimeMode),- usageCap: Schema.optional(Schema.NullOr(BotUsageCap)),imageProvider: Schema.optional(Schema.NullOr(ImageProviderId)),personalityTone: Schema.optional(BotPersonalityTone),voiceEnabled: Schema.optional(Schema.Boolean),diff --git a/packages/contracts/src/portability.ts b/packages/contracts/src/portability.tsindex 4f5d1d200..09109c282 100644--- a/packages/contracts/src/portability.ts+++ b/packages/contracts/src/portability.ts@@ -20,7 +20,6 @@ import {BotPersonalityTone,BotSandboxBrowserSharing,PersistedBotSandbox,- BotUsageCap,GroupBotMembership,} from "./orchestration/roster.ts";import {@@ -98,7 +97,14 @@ export const PortabilityBotData = Schema.Struct({engine: Schema.NullOr(BotEngine),sandbox: PersistedBotSandbox,runtimeMode: RuntimeMode,- usageCap: Schema.NullOr(BotUsageCap),+ /**+ * The retired token hard stop. Archives exported before its removal still+ * carry it and sign it, so it decodes as-is to keep their checksums valid;+ * import ignores it and new exports omit it.+ */+ usageCap: Schema.optionalKey(+ Schema.NullOr(Schema.Struct({ unit: Schema.Literal("tokens"), limit: Schema.Number })),+ ),imageProvider: Schema.NullOr(ImageProviderId).pipe(Schema.withDecodingDefault(Effect.succeed(null)),),diff --git a/packages/shared/src/model.test.ts b/packages/shared/src/model.test.tsindex 61727e2ed..c82790abc 100644--- a/packages/shared/src/model.test.ts+++ b/packages/shared/src/model.test.ts@@ -3,6 +3,8 @@ import { ProviderDriverKind, ProviderInstanceId, type ModelCapabilities } from "import {applyClaudePromptEffortPrefix,+ botEngineFromModelSelection,+ botEngineModelSelection,buildProviderOptionSelectionsFromDescriptors,createModelCapabilities,createModelSelection,@@ -13,6 +15,8 @@ import {getProviderOptionStringSelectionValue,normalizeCustomModelSlug,normalizeModelSlug,+ providerOptionsForModelChange,+ retainChosenProviderOptions,} from "./model.ts";const codexCaps: ModelCapabilities = createModelCapabilities({@@ -186,3 +190,209 @@ describe("applyClaudePromptEffortPrefix", () => {);});});++const nativeReasoningCaps = (levels: ReadonlyArray<string>): ModelCapabilities =>+ createModelCapabilities({+ optionDescriptors: [+ {+ id: "reasoningEffort",+ label: "Reasoning",+ type: "select",+ currentValue: "default",+ options: [+ { id: "default", label: "Provider default", isDefault: true },+ ...levels.map((id) => ({ id, label: id })),+ ],+ },+ ],+ });++const haikuCaps: ModelCapabilities = createModelCapabilities({+ optionDescriptors: [{ id: "thinking", label: "Thinking", type: "boolean" }],+});++describe("bot engine option helpers", () => {+ it("shows and retains legacy off as none only when the model offers none", () => {+ const options = [{ id: "reasoningEffort", value: "off" }];+ const caps = nativeReasoningCaps(["none", "low"]);+ const descriptors = getProviderOptionDescriptors({ caps, selections: options });++ expect(descriptors[0]?.currentValue).toBe("none");+ expect(buildProviderOptionSelectionsFromDescriptors(descriptors)).toEqual([+ { id: "reasoningEffort", value: "none" },+ ]);+ expect(retainChosenProviderOptions(options, caps)).toEqual([+ { id: "reasoningEffort", value: "none" },+ ]);+ expect(+ providerOptionsForModelChange({+ previous: createModelSelection(ProviderInstanceId.make("codex"), "gpt-a", options),+ instanceId: "codex",+ nextCaps: caps,+ }),+ ).toEqual([{ id: "reasoningEffort", value: "none" }]);+ expect(retainChosenProviderOptions(options, nativeReasoningCaps(["low"]))).toBeUndefined();+ expect(retainChosenProviderOptions(options, nativeReasoningCaps(["off"]))).toEqual(options);+ });++ it("keeps only advertised, non-default choices", () => {+ expect(+ retainChosenProviderOptions(+ [+ { id: "reasoningEffort", value: "minimal" },+ { id: "serviceTier", value: "fast" },+ ],+ nativeReasoningCaps(["minimal", "high"]),+ ),+ ).toEqual([{ id: "reasoningEffort", value: "minimal" }]);+ });++ it("removes a choice reset to the provider default", () => {+ expect(+ retainChosenProviderOptions(+ [{ id: "reasoningEffort", value: "default" }],+ nativeReasoningCaps(["high"]),+ ),+ ).toBeUndefined();+ });++ it("keeps a real choice whose id is spelled default when it is not the default", () => {+ const caps = createModelCapabilities({+ optionDescriptors: [+ {+ id: "variant",+ label: "Variant",+ type: "select",+ options: [+ { id: "high", label: "High", isDefault: true },+ { id: "default", label: "Default" },+ ],+ },+ ],+ });++ expect(retainChosenProviderOptions([{ id: "variant", value: "default" }], caps)).toEqual([+ { id: "variant", value: "default" },+ ]);+ expect(retainChosenProviderOptions([{ id: "variant", value: "high" }], caps)).toBeUndefined();+ });++ it("drops prompt-injected workflow values and unsupported levels", () => {+ expect(+ retainChosenProviderOptions(+ [+ { id: "effort", value: "ultrathink" },+ { id: "contextWindow", value: "200k" },+ ],+ claudeCaps,+ ),+ ).toEqual([{ id: "contextWindow", value: "200k" }]);+ expect(+ retainChosenProviderOptions(+ [{ id: "reasoningEffort", value: "max" }],+ nativeReasoningCaps(["high"]),+ ),+ ).toBeUndefined();+ });++ it("preserves both values of a boolean without an advertised default", () => {+ expect(retainChosenProviderOptions([{ id: "thinking", value: false }], haikuCaps)).toEqual([+ { id: "thinking", value: false },+ ]);+ expect(retainChosenProviderOptions([{ id: "thinking", value: true }], haikuCaps)).toEqual([+ { id: "thinking", value: true },+ ]);+ });++ it("keeps options when capabilities are unknown but clears them for an authoritative empty list", () => {+ const options = [{ id: "reasoningEffort", value: "high" }];++ expect(retainChosenProviderOptions(options, null)).toEqual(options);+ expect(retainChosenProviderOptions(options, {})).toEqual(options);+ expect(+ retainChosenProviderOptions(options, createModelCapabilities({ optionDescriptors: [] })),+ ).toBeUndefined();+ });++ it("carries supported choices to another model on the same instance", () => {+ const previous = createModelSelection(ProviderInstanceId.make("codex"), "gpt-a", [+ { id: "reasoningEffort", value: "minimal" },+ ]);++ expect(+ providerOptionsForModelChange({+ previous,+ instanceId: "codex",+ nextCaps: nativeReasoningCaps(["minimal", "high"]),+ }),+ ).toEqual([{ id: "reasoningEffort", value: "minimal" }]);+ expect(+ providerOptionsForModelChange({+ previous,+ instanceId: "codex",+ nextCaps: nativeReasoningCaps(["high"]),+ }),+ ).toBeUndefined();+ expect(+ providerOptionsForModelChange({ previous, instanceId: "codex", nextCaps: null }),+ ).toBeUndefined();+ });++ it("clears choices when the model moves to another provider instance", () => {+ expect(+ providerOptionsForModelChange({+ previous: createModelSelection(ProviderInstanceId.make("codex"), "gpt-a", [+ { id: "reasoningEffort", value: "high" },+ ]),+ instanceId: "grok",+ nextCaps: nativeReasoningCaps(["high"]),+ }),+ ).toBeUndefined();+ });++ it("reads a saved engine without borrowing anything and saves only chosen options", () => {+ expect(botEngineModelSelection({ provider: "claudeAgent", model: "claude-opus-5-5" })).toEqual({+ instanceId: "claudeAgent",+ model: "claude-opus-5-5",+ });+ expect(+ botEngineFromModelSelection(+ createModelSelection(ProviderInstanceId.make("codex"), "gpt-a", [+ { id: "reasoningEffort", value: "default" },+ ]),+ nativeReasoningCaps(["high"]),+ ),+ ).toEqual({ provider: "codex", model: "gpt-a" });+ });++ it("keeps an explicit OpenCode variant that matches a common default", () => {+ const caps = {+ optionDescriptors: [+ {+ id: "variant",+ label: "Variant",+ type: "select" as const,+ currentValue: "default",+ options: [+ { id: "default", label: "Provider default", isDefault: true },+ { id: "medium", label: "Medium" },+ { id: "high", label: "High" },+ ],+ },+ ],+ };++ expect(+ botEngineFromModelSelection(+ createModelSelection(ProviderInstanceId.make("opencode"), "openai/gpt-5.4", [+ { id: "variant", value: "medium" },+ ]),+ caps,+ ),+ ).toEqual({+ provider: "opencode",+ model: "openai/gpt-5.4",+ options: [{ id: "variant", value: "medium" }],+ });+ });+});diff --git a/packages/shared/src/model.ts b/packages/shared/src/model.tsindex 2ca5bcec6..6c21373b5 100644--- a/packages/shared/src/model.ts+++ b/packages/shared/src/model.ts@@ -1,5 +1,6 @@import * as Predicate from "effect/Predicate";import {+ type BotEngine,DEFAULT_MODEL,DEFAULT_MODEL_BY_PROVIDER,MODEL_SLUG_ALIASES_BY_PROVIDER,@@ -81,21 +82,36 @@ export function getModelSelectionBooleanOptionValue(return getProviderOptionBooleanSelectionValue(modelSelection?.options, id);}+function normalizeDescriptorSelectionValue(+ descriptor: ProviderOptionDescriptor,+ value: string | boolean,+): string | boolean {+ return descriptor.type === "select" &&+ descriptor.id === "reasoningEffort" &&+ value === "off" &&+ !descriptor.options.some((option) => option.id === "off") &&+ descriptor.options.some((option) => option.id === "none")+ ? "none"+ : value;+}+function resolveDescriptorChoiceValue(descriptor: Extract<ProviderOptionDescriptor, { type: "select" }>,raw: string | null | undefined,): string | undefined {- const trimmed = trimOrNull(raw);+ const trimmedRaw = trimOrNull(raw);+ const trimmed = trimmedRaw ? normalizeDescriptorSelectionValue(descriptor, trimmedRaw) : null;if (!trimmed) {return descriptor.currentValue ?? descriptor.options.find((option) => option.isDefault)?.id;}if (descriptor.options.length === 0) {- return trimmed;+ return Predicate.isString(trimmed) ? trimmed : undefined;}if (+ Predicate.isString(trimmed) &&descriptor.promptInjectedValues?.includes(trimmed) &&descriptor.options.some((option) => option.id === trimmed)) {@@ -103,7 +119,7 @@ function resolveDescriptorChoiceValue(}if (descriptor.options.some((option) => option.id === trimmed)) {- return trimmed;+ return Predicate.isString(trimmed) ? trimmed : undefined;}return descriptor.currentValue ?? descriptor.options.find((option) => option.isDefault)?.id;@@ -367,6 +383,99 @@ export function createModelSelection(return selections.length > 0 ? { ...base, options: selections } : base;}+function isChosenProviderOption(+ descriptor: ProviderOptionDescriptor,+ value: string | boolean,+): boolean {+ if (descriptor.type === "boolean") {+ return Predicate.isBoolean(value) && value !== descriptor.currentValue;+ }++ return (+ Predicate.isString(value) &&+ descriptor.options.some((option) => option.id === value && option.isDefault !== true) &&+ !descriptor.promptInjectedValues?.includes(value)+ );+}++/**+ * The options a user actually chose for a model: each names an advertised+ * descriptor and choice and differs from that descriptor's default, so picking+ * the default removes the option. Capabilities that are unknown (`null`, or a+ * snapshot without a descriptor list) cannot reject anything, so the options+ * are kept as they are. Returns undefined when nothing remains.+ */+export function retainChosenProviderOptions(+ options: ReadonlyArray<ProviderOptionSelection> | null | undefined,+ caps: ModelCapabilities | null | undefined,+): Array<ProviderOptionSelection> | undefined {+ if (!options || options.length === 0) return undefined;++ const descriptors = caps?.optionDescriptors;++ if (descriptors === undefined) return cloneSelections(options);++ const seen = new Set<string>();++ const kept = options.flatMap((option) => {+ if (seen.has(option.id)) return [];+ seen.add(option.id);+ const descriptor = descriptors.find((candidate) => candidate.id === option.id);++ if (!descriptor) return [];++ const value = normalizeDescriptorSelectionValue(descriptor, option.value);++ return isChosenProviderOption(descriptor, value) ? [{ id: option.id, value }] : [];+ });++ return kept.length > 0 ? cloneSelections(kept) : undefined;+}++/**+ * The options a model pick carries over from `previous`. Another provider+ * instance starts clean. The same instance keeps the choices the next model+ * supports; a next model with unknown capabilities starts clean, because+ * nothing confirms the old choices apply to it.+ */+export function providerOptionsForModelChange(input: {+ readonly previous: ModelSelection | null | undefined;+ readonly instanceId: string;+ readonly nextCaps: ModelCapabilities | null | undefined;+}): Array<ProviderOptionSelection> | undefined {+ if (!input.previous || input.previous.instanceId !== input.instanceId) return undefined;++ if (input.nextCaps?.optionDescriptors === undefined) return undefined;++ return retainChosenProviderOptions(input.previous.options, input.nextCaps);+}++/**+ * A bot's saved engine as the selection it runs with. The saved options are+ * the bot's own: the app's default selection never fills them in.+ */+export function botEngineModelSelection(engine: BotEngine): ModelSelection {+ return createModelSelection(+ ProviderInstanceId.make(engine.provider),+ engine.model,+ engine.options,+ );+}++/** The engine to save for a selection, keeping only options the user chose. */+export function botEngineFromModelSelection(+ selection: ModelSelection,+ caps: ModelCapabilities | null | undefined,+): BotEngine {+ const options = retainChosenProviderOptions(selection.options, caps);++ return {+ provider: selection.instanceId,+ model: selection.model,+ ...(options ? { options } : {}),+ };+}+/*** Returns the effort value if it is a prompt-injected value according to* any select descriptor in the given capabilities, or null otherwise.