Skip to content

Never stop coding. Free MIT AI gateway: one endpoint, 375 providers (150+ free), 1,400+ models Kimi, Claude, GPT, Gemini, GLM, DeepSeek, MiniMax. Works with Claude Code, Codex, Cursor, OpenCode, Cline & Copilot. Quota-aware auto-fallback, RTK+Caveman compression saves 15-95% tokens, MCP/A2A, Desktop/PWA. Built by 900+ contributors

feat(connections): per-connection disable-cooldown opt-out (#2997) (#3852)

Diego Rodrigues de Sa e Souza7c080941d1e3
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 7d8257096d..0a283a411f 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -9,6 +9,7 @@
### 🐛 Fixed
- **fix(intelligence): run pricing + models.dev sync from the live startup path** — like the Arena ELO sync (v3.8.24), the external **pricing sync** (`PRICING_SYNC_ENABLED`) and the **models.dev capability sync** (Settings → AI toggle) were only initialized from `server-init.ts`, which the Next standalone runtime never executes — and models.dev had no caller at all. Their toggles were inert in production. Both are now initialized from `instrumentation-node.ts` (self-gated, opt-in preserved, non-blocking, never fatal). (thanks @diegosouzapw)
+- **feat(connections): per-connection "disable cooldown" opt-out** — a connection can now opt out of the transient cooldown (`providerSpecificData.disableCooling`, with a toggle in the Edit Connection modal). When set, a recoverable failure still records the error/backoff but does **not** take the connection out of rotation, so it stays eligible for selection — useful for a primary key you never want parked on a blip. Terminal states (banned / expired / credits_exhausted) still apply. ([#2997](https://github.com/diegosouzapw/OmniRoute/issues/2997) — thanks @diegosouzapw)
- **fix(combo): restore sessionless combo stickiness + reasoning-aware readiness (504 / TPS regression after v3.8.14)** — #3399 (v3.8.16) replaced the `<omniModel>`-tag combo pinning with a server-side context-cache pin gated on a client `sessionId`. Clients that send no session id (most OpenAI-compatible tools) lost combo stickiness, so combos re-ran strategy selection every turn → upstream prompt-cache misses → cold high-reasoning starts (~78s) → intermittent `[504] Upstream request did not return response headers` + TPS collapse (only on combos). The pin now falls back to a stable per-conversation fingerprint (`extractSessionAffinityKey(body)`) when no session id is present — **only when `context_cache_protection` is on**, so #3399's anti-leak behaviour is preserved. Separately, the stream-readiness window now grants the +30s reasoning budget **unconditionally** for high-reasoning Codex GPT-5.x (small high-reasoning prompts were 504-ing at the 80s base regardless of stickiness). ([#3825](https://github.com/diegosouzapw/OmniRoute/issues/3825) — thanks @bypanghu)
- **test(combo): cover the `skipProviderBreaker` consumer gate** — the producer was tested but the consumer (whether a failed combo target trips the whole-provider circuit breaker) was not; the breaker decision is now an exported pure predicate (`shouldRecordProviderBreakerFailure`, behaviour-identical) with direct tests asserting a `connection_cooldown` 503 does not trip the breaker while a plain 503 does. Closes another deferred test gap from [#2743](https://github.com/diegosouzapw/OmniRoute/issues/2743). (thanks @diegosouzapw)
- **fix(providers): surface the real Devin error + correct the Windsurf auth instructions** — Devin chat returned a generic 502 "Invalid SSE response for non-streaming request" that swallowed the real cause (e.g. "Devin CLI not found"): an error-only SSE chunk (no `choices`) is now propagated with its sanitized message. The Windsurf "Visit windsurf.com/show-auth-token" instruction (the bare URL shows no token without an IDE-supplied `?state=`) now directs users to the `Windsurf: Provide Auth Token` command-palette flow. ([#3324](https://github.com/diegosouzapw/OmniRoute/issues/3324) — thanks @mikmaneggahommie)
diff --git a/file-size-baseline.json b/file-size-baseline.json
index 1121f71073..ec7a770e43 100644
--- a/file-size-baseline.json
+++ b/file-size-baseline.json
@@ -52,7 +52,7 @@
"src/app/(dashboard)/dashboard/providers/[id]/ProviderDetailPageClient.tsx": 784,
"src/app/(dashboard)/dashboard/providers/[id]/components/ConnectionRow.tsx": 941,
"src/app/(dashboard)/dashboard/providers/[id]/components/modals/AddApiKeyModal.tsx": 843,
- "src/app/(dashboard)/dashboard/providers/[id]/components/modals/EditConnectionModal.tsx": 1171,
+ "src/app/(dashboard)/dashboard/providers/[id]/components/modals/EditConnectionModal.tsx": 1174,
"src/app/(dashboard)/dashboard/providers/[id]/hooks/useProviderConnections.ts": 954,
"src/app/(dashboard)/dashboard/providers/[id]/hooks/useProviderModels.ts": 155,
"src/app/(dashboard)/dashboard/providers/[id]/hooks/useProviderSettings.ts": 264,
@@ -103,7 +103,7 @@
"src/shared/services/cliRuntime.ts": 1090,
"src/shared/validation/schemas.ts": 2523,
"src/sse/handlers/chat.ts": 1425,
- "src/sse/services/auth.ts": 2207
+ "src/sse/services/auth.ts": 2216
},
"_rebaseline_2026_06_09": "Re-baseline consciente pre-release v3.8.19: 9 arquivos cresceram durante o ciclo (features mergeadas: RequestLoggerV2 +281 request-logger rework, stream +101, combo +73, chatCore +45, catalog +32 fable-5/catalog-flag, callLogs +4, accountFallback +2, usageHistory novo 840) + core.ts +7 (fix resetAllDbModuleState, PR 3536). A catraca segue valendo destes valores — proximo crescimento falha. Decisao: encolher (esp. RequestLoggerV2/chatCore) e a issue #3501 ficam para o ciclo seguinte.",
"_rebaseline_2026_06_11_phase1f": "Phase 1f (#3501): ProviderDetailPageClient.tsx 4948→4062 (-886 LOC); 3 novos hooks extraídos. useProviderConnections.ts=954 acima do cap=800 — justificado: extração direta do god-component (zero lógica nova), própria redução do cliente supera o custo. useProviderSettings.ts=263 e useProviderModels.ts=154 já abaixo do cap.",
@@ -128,5 +128,6 @@
"_rebaseline_2026_06_14_r3_3849_transient_hide": "PR #3849 own growth: providerPageHelpers.ts 939→955 (+16 = expanded JSDoc on the auto-hide policy + transient-failure guard in evaluateTestAllEntry). Cohesive helper logic; not extractable.",
"_rebaseline_2026_06_14_r3_3838_opencode_quota": "PR #3838 own growth: usage.ts 3394→3408 (+14 = clearer OpenCode Go missing-quota-API messages with OMNIROUTE_OPENCODE_GO_QUOTA_URL override hint + upstream issue links). Message text only; no logic extractable.",
"_rebaseline_2026_06_14_r3_3839_veo_video": "PR #3839 own growth: schemas.ts 2522→2523 (+1 = Veo video model (predictLongRunning) validation for Gemini/Vertex dynamic discovery).",
- "_rebaseline_2026_06_14_r3_3836_kiro_discovery": "PR #3836 own growth: models/route.ts 2426→2487 (+61 = kiro live per-account discovery branch wiring fetchKiroAvailableModels into the existing cache/auto-fetch/fallback discovery flow). Structural shrink of this route tracked in #3789."
+ "_rebaseline_2026_06_14_r3_3836_kiro_discovery": "PR #3836 own growth: models/route.ts 2426→2487 (+61 = kiro live per-account discovery branch wiring fetchKiroAvailableModels into the existing cache/auto-fetch/fallback discovery flow). Structural shrink of this route tracked in #3789.",
+ "_rebaseline_2026_06_14_2997_disable_cooling": "Re-baseline #2997 (per-connection disable-cooling): EditConnectionModal.tsx 1171→1174 (+toggle UI) + auth.ts 2207→2216 (honor de disableCooling no markAccountUnavailable, pós-prettier). Lógica coesa; não-extraível. (combo.ts/RequestLoggerV2 drift já documentado em _r3_3835.)"
}
diff --git a/open-sse/services/kiroModels.ts b/open-sse/services/kiroModels.ts
index 5e1a25a3ac..24bc13d810 100644
--- a/open-sse/services/kiroModels.ts
+++ b/open-sse/services/kiroModels.ts
@@ -1,196 +1,195 @@
-/**
- * Kiro (AWS CodeWhisperer / Amazon Q) live model discovery.
- *
- * Kiro's model catalog is per-account / per-tier — the free tier, Pro, Pro+ and
- * Power plans expose different model sets, and AWS IAM Identity Center (enterprise)
- * orgs further restrict it to an admin-curated "approved models" list. The Kiro
- * IDE / CLI populates its model picker by calling the CodeWhisperer
- * `ListAvailableModels` operation:
- *
- * GET https://q.{region}.amazonaws.com/ListAvailableModels?origin=AI_EDITOR
- * Authorization: Bearer <accessToken>
- * → { models: [ { modelId, modelName?, tokenLimits?: { maxInputTokens } }, ... ] }
- *
- * This works for both "simple" Builder ID / social logins and AWS IAM Identity
- * Center accounts:
- * - `origin=AI_EDITOR` alone is the universal call (Builder ID / IdC).
- * - `profileArn` is only sent for desktop-style accounts that have one, and only
- * as a retry, because sending it for Builder ID can yield 403.
- * - The endpoint is region-matched (IdC tokens are region-bound, e.g.
- * eu-central-1) with a us-east-1 fallback (the legacy CodeWhisperer home region).
- *
- * A safe fallback to the static registry catalog is preserved so model import
- * never breaks when the account is offline / unauthenticated / token-expired.
- */
-
-type RawRecord = Record<string, unknown>;
-
-function asRecord(value: unknown): RawRecord {
- return value && typeof value === "object" && !Array.isArray(value) ? (value as RawRecord) : {};
-}
-
-function toNonEmptyString(value: unknown): string | null {
- if (typeof value !== "string") return null;
- const trimmed = value.trim();
- return trimmed.length > 0 ? trimmed : null;
-}
-
-export type KiroModel = {
- id: string;
- name: string;
- owned_by: string;
-};
-
-export type KiroModelsResult = {
- models: KiroModel[];
- /** "api" = live discovery; "fallback" = static catalog (offline/unauthed/error). */
- source: "api" | "fallback";
-};
-
-/**
- * Parse a CodeWhisperer `ListAvailableModels` response into managed model rows.
- * Only ids present in the live response are returned, which gives the exact
- * per-account / per-tier entitlement filtering.
- */
-export function parseKiroModels(data: unknown): KiroModel[] {
- const payload = asRecord(data);
- const items = Array.isArray(payload.models)
- ? (payload.models as unknown[])
- : Array.isArray(payload.availableModels)
- ? (payload.availableModels as unknown[])
- : [];
-
- const seen = new Set<string>();
- const models: KiroModel[] = [];
-
- for (const value of items) {
- const item = asRecord(value);
- const id = toNonEmptyString(item.modelId) || toNonEmptyString(item.id);
- if (!id || seen.has(id)) continue;
- seen.add(id);
- const name =
- toNonEmptyString(item.modelName) || toNonEmptyString(item.name) || id;
- models.push({ id, name, owned_by: "kiro" });
- }
-
- return models;
-}
-
-/**
- * Derive the AWS region for a Kiro connection. Mirrors getKiroUsage: prefer the
- * stored region, then the region embedded in the profileArn, else us-east-1.
- */
-export function resolveKiroRegion(providerSpecificData: unknown): string {
- const psd = asRecord(providerSpecificData);
- const explicit = toNonEmptyString(psd.region);
- if (explicit) return explicit.toLowerCase();
-
- const profileArn = toNonEmptyString(psd.profileArn);
- const fromArn = profileArn
- ? profileArn.toLowerCase().match(/^arn:aws:codewhisperer:([a-z0-9-]+):/)?.[1]
- : undefined;
-
- return fromArn || "us-east-1";
-}
-
-/**
- * Build the ordered list of `ListAvailableModels` base URLs to try: the
- * region-matched Amazon Q host first, then the us-east-1 home region as a
- * fallback (CodeWhisperer's canonical region).
- */
-export function buildKiroModelsEndpoints(region: string): string[] {
- const normalized = (toNonEmptyString(region) || "us-east-1").toLowerCase();
- const urls: string[] = [`https://q.${normalized}.amazonaws.com/ListAvailableModels`];
- if (normalized !== "us-east-1") {
- urls.push("https://q.us-east-1.amazonaws.com/ListAvailableModels");
- }
- return urls;
-}
-
-export type FetchKiroModelsOptions = {
- /** Stored Kiro access token (Bearer). */
- accessToken: string | null | undefined;
- /** Connection providerSpecificData (region, profileArn). */
- providerSpecificData?: unknown;
- /** Injectable fetch (defaults to global fetch). */
- fetchImpl?: typeof fetch;
- /** Static catalog to fall back to when live discovery is unavailable. */
- fallbackModels?: Array<{ id: string; name?: string }>;
-};
-
-function toFallbackResult(
- fallbackModels: Array<{ id: string; name?: string }> | undefined
-): KiroModelsResult {
- const models = (fallbackModels || [])
- .map((model) => {
- const id = toNonEmptyString(model.id);
- if (!id) return null;
- return { id, name: toNonEmptyString(model.name) || id, owned_by: "kiro" };
- })
- .filter((model): model is KiroModel => Boolean(model));
- return { models, source: "fallback" };
-}
-
-async function tryFetchModels(
- fetchImpl: typeof fetch,
- url: string,
- accessToken: string
-): Promise<KiroModel[] | null> {
- try {
- const response = await fetchImpl(url, {
- method: "GET",
- headers: {
- Authorization: `Bearer ${accessToken}`,
- Accept: "application/json",
- },
- });
- if (!response.ok) return null;
- const data = await response.json();
- const models = parseKiroModels(data);
- return models.length > 0 ? models : null;
- } catch {
- return null;
- }
-}
-
-/**
- * Discover the Kiro model catalog live via `ListAvailableModels`, falling back
- * to the static catalog when no token is available or every attempt fails.
- *
- * Attempt order (stops at the first success):
- * 1. `origin=AI_EDITOR` on each region-matched endpoint — universal path that
- * works for Builder ID / social ("simple") and IAM Identity Center accounts.
- * 2. `origin=AI_EDITOR&profileArn=...` on the primary endpoint, only when a
- * profileArn is present (desktop-style accounts that require it).
- */
-export async function fetchKiroAvailableModels(
- options: FetchKiroModelsOptions
-): Promise<KiroModelsResult> {
- const { accessToken, providerSpecificData, fetchImpl = fetch, fallbackModels } = options;
-
- const token = toNonEmptyString(accessToken);
- if (!token) {
- return toFallbackResult(fallbackModels);
- }
-
- const region = resolveKiroRegion(providerSpecificData);
- const endpoints = buildKiroModelsEndpoints(region);
- const profileArn = toNonEmptyString(asRecord(providerSpecificData).profileArn);
-
- // Pass 1: origin-only (works for Builder ID / social / IdC).
- for (const base of endpoints) {
- const models = await tryFetchModels(fetchImpl, `${base}?origin=AI_EDITOR`, token);
- if (models) return { models, source: "api" };
- }
-
- // Pass 2: retry with profileArn (desktop accounts that require it) on the
- // region-matched endpoint only. Skipped for Builder ID / IdC where sending a
- // profileArn can 403.
- if (profileArn) {
- const url = `${endpoints[0]}?origin=AI_EDITOR&profileArn=${encodeURIComponent(profileArn)}`;
- const models = await tryFetchModels(fetchImpl, url, token);
- if (models) return { models, source: "api" };
- }
-
- return toFallbackResult(fallbackModels);
-}
+/**
+ * Kiro (AWS CodeWhisperer / Amazon Q) live model discovery.
+ *
+ * Kiro's model catalog is per-account / per-tier — the free tier, Pro, Pro+ and
+ * Power plans expose different model sets, and AWS IAM Identity Center (enterprise)
+ * orgs further restrict it to an admin-curated "approved models" list. The Kiro
+ * IDE / CLI populates its model picker by calling the CodeWhisperer
+ * `ListAvailableModels` operation:
+ *
+ * GET https://q.{region}.amazonaws.com/ListAvailableModels?origin=AI_EDITOR
+ * Authorization: Bearer <accessToken>
+ * → { models: [ { modelId, modelName?, tokenLimits?: { maxInputTokens } }, ... ] }
+ *
+ * This works for both "simple" Builder ID / social logins and AWS IAM Identity
+ * Center accounts:
+ * - `origin=AI_EDITOR` alone is the universal call (Builder ID / IdC).
+ * - `profileArn` is only sent for desktop-style accounts that have one, and only
+ * as a retry, because sending it for Builder ID can yield 403.
+ * - The endpoint is region-matched (IdC tokens are region-bound, e.g.
+ * eu-central-1) with a us-east-1 fallback (the legacy CodeWhisperer home region).
+ *
+ * A safe fallback to the static registry catalog is preserved so model import
+ * never breaks when the account is offline / unauthenticated / token-expired.
+ */
+
+type RawRecord = Record<string, unknown>;
+
+function asRecord(value: unknown): RawRecord {
+ return value && typeof value === "object" && !Array.isArray(value) ? (value as RawRecord) : {};
+}
+
+function toNonEmptyString(value: unknown): string | null {
+ if (typeof value !== "string") return null;
+ const trimmed = value.trim();
+ return trimmed.length > 0 ? trimmed : null;
+}
+
+export type KiroModel = {
+ id: string;
+ name: string;
+ owned_by: string;
+};
+
+export type KiroModelsResult = {
+ models: KiroModel[];
+ /** "api" = live discovery; "fallback" = static catalog (offline/unauthed/error). */
+ source: "api" | "fallback";
+};
+
+/**
+ * Parse a CodeWhisperer `ListAvailableModels` response into managed model rows.
+ * Only ids present in the live response are returned, which gives the exact
+ * per-account / per-tier entitlement filtering.
+ */
+export function parseKiroModels(data: unknown): KiroModel[] {
+ const payload = asRecord(data);
+ const items = Array.isArray(payload.models)
+ ? (payload.models as unknown[])
+ : Array.isArray(payload.availableModels)
+ ? (payload.availableModels as unknown[])
+ : [];
+
+ const seen = new Set<string>();
+ const models: KiroModel[] = [];
+
+ for (const value of items) {
+ const item = asRecord(value);
+ const id = toNonEmptyString(item.modelId) || toNonEmptyString(item.id);
+ if (!id || seen.has(id)) continue;
+ seen.add(id);
+ const name = toNonEmptyString(item.modelName) || toNonEmptyString(item.name) || id;
+ models.push({ id, name, owned_by: "kiro" });
+ }
+
+ return models;
+}
+
+/**
+ * Derive the AWS region for a Kiro connection. Mirrors getKiroUsage: prefer the
+ * stored region, then the region embedded in the profileArn, else us-east-1.
+ */
+export function resolveKiroRegion(providerSpecificData: unknown): string {
+ const psd = asRecord(providerSpecificData);
+ const explicit = toNonEmptyString(psd.region);
+ if (explicit) return explicit.toLowerCase();
+
+ const profileArn = toNonEmptyString(psd.profileArn);
+ const fromArn = profileArn
+ ? profileArn.toLowerCase().match(/^arn:aws:codewhisperer:([a-z0-9-]+):/)?.[1]
+ : undefined;
+
+ return fromArn || "us-east-1";
+}
+
+/**
+ * Build the ordered list of `ListAvailableModels` base URLs to try: the
+ * region-matched Amazon Q host first, then the us-east-1 home region as a
+ * fallback (CodeWhisperer's canonical region).
+ */
+export function buildKiroModelsEndpoints(region: string): string[] {
+ const normalized = (toNonEmptyString(region) || "us-east-1").toLowerCase();
+ const urls: string[] = [`https://q.${normalized}.amazonaws.com/ListAvailableModels`];
+ if (normalized !== "us-east-1") {
+ urls.push("https://q.us-east-1.amazonaws.com/ListAvailableModels");
+ }
+ return urls;
+}
+
+export type FetchKiroModelsOptions = {
+ /** Stored Kiro access token (Bearer). */
+ accessToken: string | null | undefined;
+ /** Connection providerSpecificData (region, profileArn). */
+ providerSpecificData?: unknown;
+ /** Injectable fetch (defaults to global fetch). */
+ fetchImpl?: typeof fetch;
+ /** Static catalog to fall back to when live discovery is unavailable. */
+ fallbackModels?: Array<{ id: string; name?: string }>;
+};
+
+function toFallbackResult(
+ fallbackModels: Array<{ id: string; name?: string }> | undefined
+): KiroModelsResult {
+ const models = (fallbackModels || [])
+ .map((model) => {
+ const id = toNonEmptyString(model.id);
+ if (!id) return null;
+ return { id, name: toNonEmptyString(model.name) || id, owned_by: "kiro" };
+ })
+ .filter((model): model is KiroModel => Boolean(model));
+ return { models, source: "fallback" };
+}
+
+async function tryFetchModels(
+ fetchImpl: typeof fetch,
+ url: string,
+ accessToken: string
+): Promise<KiroModel[] | null> {
+ try {
+ const response = await fetchImpl(url, {
+ method: "GET",
+ headers: {
+ Authorization: `Bearer ${accessToken}`,
+ Accept: "application/json",
+ },
+ });
+ if (!response.ok) return null;
+ const data = await response.json();
+ const models = parseKiroModels(data);
+ return models.length > 0 ? models : null;
+ } catch {
+ return null;
+ }
+}
+
+/**
+ * Discover the Kiro model catalog live via `ListAvailableModels`, falling back
+ * to the static catalog when no token is available or every attempt fails.
+ *
+ * Attempt order (stops at the first success):
+ * 1. `origin=AI_EDITOR` on each region-matched endpoint — universal path that
+ * works for Builder ID / social ("simple") and IAM Identity Center accounts.
+ * 2. `origin=AI_EDITOR&profileArn=...` on the primary endpoint, only when a
+ * profileArn is present (desktop-style accounts that require it).
+ */
+export async function fetchKiroAvailableModels(
+ options: FetchKiroModelsOptions
+): Promise<KiroModelsResult> {
+ const { accessToken, providerSpecificData, fetchImpl = fetch, fallbackModels } = options;
+
+ const token = toNonEmptyString(accessToken);
+ if (!token) {
+ return toFallbackResult(fallbackModels);
+ }
+
+ const region = resolveKiroRegion(providerSpecificData);
+ const endpoints = buildKiroModelsEndpoints(region);
+ const profileArn = toNonEmptyString(asRecord(providerSpecificData).profileArn);
+
+ // Pass 1: origin-only (works for Builder ID / social / IdC).
+ for (const base of endpoints) {
+ const models = await tryFetchModels(fetchImpl, `${base}?origin=AI_EDITOR`, token);
+ if (models) return { models, source: "api" };
+ }
+
+ // Pass 2: retry with profileArn (desktop accounts that require it) on the
+ // region-matched endpoint only. Skipped for Builder ID / IdC where sending a
+ // profileArn can 403.
+ if (profileArn) {
+ const url = `${endpoints[0]}?origin=AI_EDITOR&profileArn=${encodeURIComponent(profileArn)}`;
+ const models = await tryFetchModels(fetchImpl, url, token);
+ if (models) return { models, source: "api" };
+ }
+
+ return toFallbackResult(fallbackModels);
+}
diff --git a/src/app/(dashboard)/dashboard/providers/[id]/components/modals/EditConnectionModal.tsx b/src/app/(dashboard)/dashboard/providers/[id]/components/modals/EditConnectionModal.tsx
index 458483e529..5103fe1cc2 100644
--- a/src/app/(dashboard)/dashboard/providers/[id]/components/modals/EditConnectionModal.tsx
+++ b/src/app/(dashboard)/dashboard/providers/[id]/components/modals/EditConnectionModal.tsx
@@ -113,6 +113,7 @@ export default function EditConnectionModal({
? isClaudeExtraUsageBlockEnabled(connection?.provider, connection?.providerSpecificData)
: false,
passthroughModels: connection?.providerSpecificData?.passthroughModels === true,
+ disableCooling: connection?.providerSpecificData?.disableCooling === true,
});
const [testing, setTesting] = useState(false);
const [testResult, setTestResult] = useState(null);
@@ -267,6 +268,7 @@ export default function EditConnectionModal({
connection.providerSpecificData
),
passthroughModels: connection?.providerSpecificData?.passthroughModels === true,
+ disableCooling: connection?.providerSpecificData?.disableCooling === true,
});
// Load existing extra keys from providerSpecificData
const existing = connection.providerSpecificData?.extraApiKeys;
@@ -544,6 +546,11 @@ export default function EditConnectionModal({
),
};
}
+ // #2997: persist the transient-cooldown opt-out; write only when enabled,
+ // clear it otherwise so a disabled toggle does not linger as `false`.
+ if (updates.providerSpecificData) {
+ updates.providerSpecificData.disableCooling = formData.disableCooling ? true : undefined;
+ }
const error = (await onSave(updates)) as void | unknown;
if (error) {
setSaveError(typeof error === "string" ? error : t("failedSaveConnection"));
@@ -647,6 +654,15 @@ export default function EditConnectionModal({
/>
</div>
)}
+ {/* #2997: per-connection transient-cooldown opt-out (provider-agnostic) */}
+ <div className="flex flex-col gap-4 rounded-lg border border-border/50 bg-surface/20 p-4">
+ <Toggle
+ checked={formData.disableCooling}
+ onChange={(checked) => setFormData({ ...formData, disableCooling: checked })}
+ label={t("disableCoolingLabel")}
+ description={t("disableCoolingDescription")}
+ />
+ </div>
{supportsGoogleProjectId && (
<div className="flex flex-col gap-4 rounded-lg border border-border/50 bg-surface/20 p-4">
{isAntigravity && (
diff --git a/src/app/api/providers/[id]/models/route.ts b/src/app/api/providers/[id]/models/route.ts
index 3c22459d00..928a886732 100755
--- a/src/app/api/providers/[id]/models/route.ts
+++ b/src/app/api/providers/[id]/models/route.ts
@@ -78,7 +78,10 @@ import {
isAutoFetchModelsEnabled,
persistDiscoveredModels,
} from "@/lib/providerModels/modelDiscovery";
-import { parseGeminiModelsList, type GeminiDiscoveryModel } from "@/lib/providerModels/geminiModelsParser";
+import {
+ parseGeminiModelsList,
+ type GeminiDiscoveryModel,
+} from "@/lib/providerModels/geminiModelsParser";
import { getSyncedAvailableModels } from "@/lib/db/models";
import { fetchCursorAgentModels } from "@/lib/providerModels/cursorAgent";
@@ -750,8 +753,7 @@ export async function GET(
(data.data || data.models || []) as Array<Record<string, unknown>>
)
.map((item) => {
- const itemId =
- typeof item.id === "string" ? item.id.trim() : "";
+ const itemId = typeof item.id === "string" ? item.id.trim() : "";
if (!itemId) return null;
const itemName =
typeof item.display_name === "string"
@@ -845,7 +847,9 @@ export async function GET(
}
const registryCatalogModels = providerSyncedModels ?? (getModelsByProviderId(provider) || []);
- const specialtyCatalogModels = providerSyncedModels ? [] : (getStaticModelsForProvider(provider) || []);
+ const specialtyCatalogModels = providerSyncedModels
+ ? []
+ : getStaticModelsForProvider(provider) || [];
const toLocalCatalogModels = () => {
const localCatalog = mergeLocalCatalogModels(registryCatalogModels, specialtyCatalogModels);
@@ -2118,9 +2122,8 @@ export async function GET(
let queryKey: string | null = null;
let bearerToken: string | null = null;
try {
- const { parseSAFromApiKey, getAccessToken } = await import(
- "@omniroute/open-sse/executors/vertex.ts"
- );
+ const { parseSAFromApiKey, getAccessToken } =
+ await import("@omniroute/open-sse/executors/vertex.ts");
if (accessToken) {
bearerToken = accessToken;
} else if (credential) {
@@ -2130,8 +2133,7 @@ export async function GET(
let isServiceAccountJson = false;
try {
const parsed = JSON.parse(credential);
- isServiceAccountJson =
- !!parsed && typeof parsed === "object" && !Array.isArray(parsed);
+ isServiceAccountJson = !!parsed && typeof parsed === "object" && !Array.isArray(parsed);
} catch {
isServiceAccountJson = false;
}
diff --git a/src/i18n/messages/en.json b/src/i18n/messages/en.json
index 72f6bd2aa6..21ebe6a8bd 100644
--- a/src/i18n/messages/en.json
+++ b/src/i18n/messages/en.json
@@ -4438,6 +4438,8 @@
"t3ChatWebCookiePlaceholder": "convex-session-id=abc123...",
"blockClaudeExtraUsageDescription": "Hide extra Claude usage rows reported by some providers when they duplicate primary token accounting.",
"blockClaudeExtraUsageLabel": "Block duplicate Claude usage rows",
+ "disableCoolingDescription": "Skip the transient cooldown so this connection stays eligible even after recoverable errors (terminal states like banned/expired still apply).",
+ "disableCoolingLabel": "Disable cooldown for this connection",
"bulkPasteAdded": "{count, plural, one {1 key added} other {# keys added}}",
"bulkPasteDuplicatesIgnored": "{count, plural, one {1 duplicate skipped} other {# duplicates skipped}}",
"bulkPasteHint": "Paste one API key per line. Empty lines are ignored and duplicate keys are skipped.",
diff --git a/src/lib/providerModels/geminiModelsParser.ts b/src/lib/providerModels/geminiModelsParser.ts
index e4c8063d0b..2836846886 100644
--- a/src/lib/providerModels/geminiModelsParser.ts
+++ b/src/lib/providerModels/geminiModelsParser.ts
@@ -1,91 +1,91 @@
-/**
- * Parses the Google Generative Language `v1beta/models` listing into discovery models.
- *
- * Each model's `supportedGenerationMethods` is mapped to OmniRoute endpoints:
- * - generateContent / generateAnswer → "chat"
- * - predict → "images" (Imagen image generation)
- * - predictLongRunning → "video" (Veo video generation)
- * - embedContent → "embeddings"
- * - bidiGenerateContent → "audio" (Live real-time audio)
- *
- * Model-id heuristics refine the long-running bucket because Google exposes both
- * Imagen and Veo via long-running methods on the same endpoint:
- * - id contains "veo" → ensure "video"
- * - id contains "imagen" → force "images" (never "video")
- *
- * Note: `gemini-*-image` models (e.g. gemini-3-pro-image) generate images via the
- * regular `generateContent` path, so they stay "chat" (image output is a chat
- * modality) and are intentionally NOT reclassified as "images".
- *
- * This is shared by the `gemini` discovery config and the `vertex` /
- * `vertex-partner` (incl. Vertex AI Express key) discovery branches, so every
- * model the account can access — chat, image, video, audio and embeddings —
- * surfaces dynamically instead of being limited to the small static registry.
- */
-const METHOD_TO_ENDPOINT: Record<string, string> = {
- generateContent: "chat",
- embedContent: "embeddings",
- predict: "images",
- predictLongRunning: "video",
- bidiGenerateContent: "audio",
- generateAnswer: "chat",
-};
-
-const IGNORED_METHODS = new Set([
- "countTokens",
- "countTextTokens",
- "createCachedContent",
- "batchGenerateContent",
- "asyncBatchEmbedContent",
-]);
-
-export interface GeminiDiscoveryModel {
- id: string;
- name: string;
- supportedEndpoints: string[];
- inputTokenLimit?: number;
- outputTokenLimit?: number;
- description?: string;
- supportsThinking?: boolean;
- [key: string]: unknown;
-}
-
-export function parseGeminiModelsList(data: any): GeminiDiscoveryModel[] {
- return (data?.models || []).map((m: Record<string, unknown>) => {
- const methods: string[] = Array.isArray(m.supportedGenerationMethods)
- ? (m.supportedGenerationMethods as string[])
- : [];
-
- const endpoints = new Set<string>(
- methods
- .filter((method) => !IGNORED_METHODS.has(method))
- .map((method) => METHOD_TO_ENDPOINT[method] || "chat")
- );
-
- const id = ((m.name as string) || (m.id as string) || "").replace(/^models\//, "");
- const lowerId = id.toLowerCase();
-
- // Google exposes Imagen (image) and Veo (video) via long-running methods; the
- // method alone can't always distinguish them, so refine by model id.
- if (lowerId.includes("veo")) {
- endpoints.add("video");
- }
- if (lowerId.includes("imagen")) {
- endpoints.delete("video");
- endpoints.add("images");
- }
-
- if (endpoints.size === 0) endpoints.add("chat");
-
- return {
- ...m,
- id,
- name: (m.displayName as string) || id,
- supportedEndpoints: [...endpoints],
- ...(typeof m.inputTokenLimit === "number" ? { inputTokenLimit: m.inputTokenLimit } : {}),
- ...(typeof m.outputTokenLimit === "number" ? { outputTokenLimit: m.outputTokenLimit } : {}),
- ...(typeof m.description === "string" ? { description: m.description } : {}),
- ...(m.thinking === true ? { supportsThinking: true } : {}),
- } as GeminiDiscoveryModel;
- });
-}
+/**
+ * Parses the Google Generative Language `v1beta/models` listing into discovery models.
+ *
+ * Each model's `supportedGenerationMethods` is mapped to OmniRoute endpoints:
+ * - generateContent / generateAnswer → "chat"
+ * - predict → "images" (Imagen image generation)
+ * - predictLongRunning → "video" (Veo video generation)
+ * - embedContent → "embeddings"
+ * - bidiGenerateContent → "audio" (Live real-time audio)
+ *
+ * Model-id heuristics refine the long-running bucket because Google exposes both
+ * Imagen and Veo via long-running methods on the same endpoint:
+ * - id contains "veo" → ensure "video"
+ * - id contains "imagen" → force "images" (never "video")
+ *
+ * Note: `gemini-*-image` models (e.g. gemini-3-pro-image) generate images via the
+ * regular `generateContent` path, so they stay "chat" (image output is a chat
+ * modality) and are intentionally NOT reclassified as "images".
+ *
+ * This is shared by the `gemini` discovery config and the `vertex` /
+ * `vertex-partner` (incl. Vertex AI Express key) discovery branches, so every
+ * model the account can access — chat, image, video, audio and embeddings —
+ * surfaces dynamically instead of being limited to the small static registry.
+ */
+const METHOD_TO_ENDPOINT: Record<string, string> = {
+ generateContent: "chat",
+ embedContent: "embeddings",
+ predict: "images",
+ predictLongRunning: "video",
+ bidiGenerateContent: "audio",
+ generateAnswer: "chat",
+};
+
+const IGNORED_METHODS = new Set([
+ "countTokens",
+ "countTextTokens",
+ "createCachedContent",
+ "batchGenerateContent",
+ "asyncBatchEmbedContent",
+]);
+
+export interface GeminiDiscoveryModel {
+ id: string;
+ name: string;
+ supportedEndpoints: string[];
+ inputTokenLimit?: number;
+ outputTokenLimit?: number;
+ description?: string;
+ supportsThinking?: boolean;
+ [key: string]: unknown;
+}
+
+export function parseGeminiModelsList(data: any): GeminiDiscoveryModel[] {
+ return (data?.models || []).map((m: Record<string, unknown>) => {
+ const methods: string[] = Array.isArray(m.supportedGenerationMethods)
+ ? (m.supportedGenerationMethods as string[])
+ : [];
+
+ const endpoints = new Set<string>(
+ methods
+ .filter((method) => !IGNORED_METHODS.has(method))
+ .map((method) => METHOD_TO_ENDPOINT[method] || "chat")
+ );
+
+ const id = ((m.name as string) || (m.id as string) || "").replace(/^models\//, "");
+ const lowerId = id.toLowerCase();
+
+ // Google exposes Imagen (image) and Veo (video) via long-running methods; the
+ // method alone can't always distinguish them, so refine by model id.
+ if (lowerId.includes("veo")) {
+ endpoints.add("video");
+ }
+ if (lowerId.includes("imagen")) {
+ endpoints.delete("video");
+ endpoints.add("images");
+ }
+
+ if (endpoints.size === 0) endpoints.add("chat");
+
+ return {
+ ...m,
+ id,
+ name: (m.displayName as string) || id,
+ supportedEndpoints: [...endpoints],
+ ...(typeof m.inputTokenLimit === "number" ? { inputTokenLimit: m.inputTokenLimit } : {}),
+ ...(typeof m.outputTokenLimit === "number" ? { outputTokenLimit: m.outputTokenLimit } : {}),
+ ...(typeof m.description === "string" ? { description: m.description } : {}),
+ ...(m.thinking === true ? { supportsThinking: true } : {}),
+ } as GeminiDiscoveryModel;
+ });
+}
diff --git a/src/lib/providers/requestDefaults.ts b/src/lib/providers/requestDefaults.ts
index 7053956412..5f6debadff 100644
--- a/src/lib/providers/requestDefaults.ts
+++ b/src/lib/providers/requestDefaults.ts
@@ -123,6 +123,11 @@ export function normalizeProviderSpecificData(
delete normalized.blockExtraUsage;
}
+ // #2997: per-connection transient-cooldown opt-out — only persist a real boolean.
+ if ("disableCooling" in normalized && typeof normalized.disableCooling !== "boolean") {
+ delete normalized.disableCooling;
+ }
+
if ("autoFetchModels" in normalized && typeof normalized.autoFetchModels !== "boolean") {
delete normalized.autoFetchModels;
}
diff --git a/src/sse/services/auth.ts b/src/sse/services/auth.ts
index 22234e1c9e..fad736f8de 100644
--- a/src/sse/services/auth.ts
+++ b/src/sse/services/auth.ts
@@ -1790,6 +1790,15 @@ export async function markAccountUnavailable(
const connectionPassthroughModels = connProviderSpecificData.passthroughModels as
| boolean
| undefined;
+ // #2997: per-connection opt-out of the TRANSIENT connection cooldown. When set,
+ // a recoverable failure records lastError/backoff but does NOT cool the
+ // connection, so getProviderCredentials keeps selecting it. Terminal states
+ // (banned/expired/credits_exhausted) are unaffected — they are resolved below
+ // via resolveTerminalConnectionStatus() and still take the connection out.
+ // NOTE: this first cut scopes the opt-out to the CONNECTION-level cooldown only;
+ // per-model lockout branches (per-model quota 403/404, codex scope) are left
+ // as-is — extending disableCooling to model lockout is a follow-up.
+ const disableCooling = connProviderSpecificData.disableCooling === true;
const isPerModelQuotaProvider = hasPerModelQuota(provider, model, connectionPassthroughModels);
if (
@@ -1998,7 +2007,7 @@ export async function markAccountUnavailable(
await updateProviderConnection(connectionId, {
...baseUpdate,
});
- } else if (cooldownMs > 0) {
+ } else if (cooldownMs > 0 && !disableCooling) {
await updateProviderConnection(connectionId, {
...baseUpdate,
rateLimitedUntil: getUnavailableUntil(cooldownMs),
diff --git a/tests/unit/auth-disable-cooling-2997.test.ts b/tests/unit/auth-disable-cooling-2997.test.ts
new file mode 100644
index 0000000000..8c5dc95a57
--- /dev/null
+++ b/tests/unit/auth-disable-cooling-2997.test.ts
@@ -0,0 +1,191 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+
+const TEST_DATA_DIR = fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-auth-disable-cooling-"));
+process.env.DATA_DIR = TEST_DATA_DIR;
+
+const core = await import("../../src/lib/db/core.ts");
+const providersDb = await import("../../src/lib/db/providers.ts");
+const auth = await import("../../src/sse/services/auth.ts");
+
+async function resetStorage() {
+ core.resetDbInstance();
+ fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true });
+ fs.mkdirSync(TEST_DATA_DIR, { recursive: true });
+}
+
+test.after(() => {
+ core.resetDbInstance();
+ fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true });
+});
+
+// #2997 — Test 1: a recoverable error on a connection flagged disableCooling
+// must NOT write a transient cooldown (no rateLimitedUntil, testStatus stays out of
+// "unavailable"). The connection records its lastError/backoff but stays eligible.
+test("markAccountUnavailable skips transient cooldown when disableCooling is set", async () => {
+ await resetStorage();
+
+ const conn = await providersDb.createProviderConnection({
+ provider: "glm",
+ authType: "apikey",
+ apiKey: "sk-disable-cooling",
+ isActive: true,
+ testStatus: "active",
+ providerSpecificData: { disableCooling: true },
+ });
+
+ const result = await auth.markAccountUnavailable(
+ (conn as any).id,
+ 503,
+ "temporary upstream error",
+ "glm",
+ "glm-5.1"
+ );
+ const after = await providersDb.getProviderConnectionById((conn as any).id);
+
+ assert.equal(result.shouldFallback, true);
+ // No transient cooldown should be applied.
+ assert.ok(!after.rateLimitedUntil, "rateLimitedUntil must not be set when disableCooling is on");
+ assert.notEqual(
+ after.testStatus,
+ "unavailable",
+ "testStatus must not become 'unavailable' when disableCooling is on"
+ );
+ // The error is still recorded (backoff/lastError path), connection stays usable.
+ assert.equal(Number(after.errorCode), 503);
+ assert.ok(after.lastError, "lastError must still be recorded");
+});
+
+// #2997 — Test 2 (load-bearing): the flag must NEVER rescue a terminal state.
+// A 401 still expires the connection; a 402 still exhausts credits — both DESPITE
+// disableCooling being set. The guard only skips the TRANSIENT cooldown branch.
+test("markAccountUnavailable still applies terminal 'expired' despite disableCooling", async () => {
+ await resetStorage();
+
+ const conn = await providersDb.createProviderConnection({
+ provider: "openai",
+ authType: "apikey",
+ apiKey: "sk-disable-cooling-expired",
+ isActive: true,
+ testStatus: "active",
+ providerSpecificData: { disableCooling: true },
+ });
+
+ const result = await auth.markAccountUnavailable(
+ (conn as any).id,
+ 401,
+ "unauthorized",
+ "openai",
+ "gpt-4.1"
+ );
+ const after = await providersDb.getProviderConnectionById((conn as any).id);
+
+ assert.equal(result.shouldFallback, true);
+ assert.equal(after.testStatus, "expired");
+});
+
+test("markAccountUnavailable still applies terminal 'credits_exhausted' despite disableCooling", async () => {
+ await resetStorage();
+
+ const conn = await providersDb.createProviderConnection({
+ provider: "openai",
+ authType: "apikey",
+ apiKey: "sk-disable-cooling-credits",
+ isActive: true,
+ testStatus: "active",
+ providerSpecificData: { disableCooling: true },
+ });
+
+ const result = await auth.markAccountUnavailable(
+ (conn as any).id,
+ 402,
+ "payment required",
+ "openai",
+ "gpt-4.1"
+ );
+ const after = await providersDb.getProviderConnectionById((conn as any).id);
+
+ assert.equal(result.shouldFallback, true);
+ assert.equal(after.testStatus, "credits_exhausted");
+});
+
+// #2997 — Test 3: regression — a connection WITHOUT the flag keeps the default
+// transient cooldown behavior (rateLimitedUntil set, testStatus 'unavailable').
+test("markAccountUnavailable still applies transient cooldown without disableCooling", async () => {
+ await resetStorage();
+
+ const conn = await providersDb.createProviderConnection({
+ provider: "glm",
+ authType: "apikey",
+ apiKey: "sk-default-cooling",
+ isActive: true,
+ testStatus: "active",
+ });
+
+ const result = await auth.markAccountUnavailable(
+ (conn as any).id,
+ 503,
+ "temporary upstream error",
+ "glm",
+ "glm-5.1"
+ );
+ const after = await providersDb.getProviderConnectionById((conn as any).id);
+
+ assert.equal(result.shouldFallback, true);
+ assert.ok(result.cooldownMs > 0);
+ assert.ok(after.rateLimitedUntil, "rateLimitedUntil must be set for default behavior");
+ assert.equal(after.testStatus, "unavailable");
+});
+
+// #2997 — Test 4: selection eligibility — after a recoverable error, the flagged
+// connection is still selectable, while an unflagged sibling that hit the same
+// error is filtered out by its cooldown.
+test("getProviderCredentials keeps a disableCooling connection eligible after a recoverable error", async () => {
+ await resetStorage();
+
+ const flagged = await providersDb.createProviderConnection({
+ provider: "glm",
+ authType: "apikey",
+ apiKey: "sk-flagged-eligible",
+ isActive: true,
+ testStatus: "active",
+ providerSpecificData: { disableCooling: true },
+ });
+
+ const unflagged = await providersDb.createProviderConnection({
+ provider: "glm",
+ authType: "apikey",
+ apiKey: "sk-unflagged-cooled",
+ isActive: true,
+ testStatus: "active",
+ });
+
+ await auth.markAccountUnavailable(
+ (flagged as any).id,
+ 503,
+ "temporary upstream error",
+ "glm",
+ "glm-5.1"
+ );
+ await auth.markAccountUnavailable(
+ (unflagged as any).id,
+ 503,
+ "temporary upstream error",
+ "glm",
+ "glm-5.1"
+ );
+
+ // Selection must keep returning the flagged connection (never the cooled sibling).
+ for (let i = 0; i < 8; i++) {
+ const selected = await auth.getProviderCredentials("glm");
+ assert.ok(selected, "expected a selectable connection");
+ assert.equal(
+ selected.connectionId,
+ (flagged as any).id,
+ "disableCooling connection must stay eligible while the unflagged sibling is cooled"
+ );
+ }
+});
diff --git a/tests/unit/catalog-updates-v3x.test.ts b/tests/unit/catalog-updates-v3x.test.ts
index 5c7b0ea318..12489c8351 100644
--- a/tests/unit/catalog-updates-v3x.test.ts
+++ b/tests/unit/catalog-updates-v3x.test.ts
@@ -89,9 +89,10 @@ test("Every Kiro registry model resolves a non-zero pricing row (no $0.00 usage)
assert.ok(models.length > 0, "kiro must expose models");
for (const model of models) {
- const pricing = getPricingForModel("kiro", model.id) as
- | { input?: number; output?: number }
- | null;
+ const pricing = getPricingForModel("kiro", model.id) as {
+ input?: number;
+ output?: number;
+ } | null;
assert.ok(pricing, `kiro pricing must include "${model.id}"`);
assert.equal(
typeof pricing?.input === "number" && typeof pricing?.output === "number",
diff --git a/tests/unit/gemini-models-parser.test.ts b/tests/unit/gemini-models-parser.test.ts
index 79b6ddf216..83007103c5 100644
--- a/tests/unit/gemini-models-parser.test.ts
+++ b/tests/unit/gemini-models-parser.test.ts
@@ -1,112 +1,110 @@
-import test from "node:test";
-import assert from "node:assert/strict";
-
-import { parseGeminiModelsList } from "../../src/lib/providerModels/geminiModelsParser";
-
-// A representative slice of the live generativelanguage v1beta/models response — including the
-// image models (gemini-*-image via generateContent, imagen-* via predict) that the Vertex catalog
-// must surface dynamically.
-const SAMPLE = {
- models: [
- {
- name: "models/gemini-2.5-flash",
- displayName: "Gemini 2.5 Flash",
- inputTokenLimit: 1048576,
- outputTokenLimit: 65536,
- supportedGenerationMethods: ["generateContent", "countTokens", "batchGenerateContent"],
- thinking: true,
- },
- {
- name: "models/gemini-3-pro-image-preview",
- displayName: "Gemini 3 Pro Image Preview",
- supportedGenerationMethods: ["generateContent", "countTokens"],
- },
- {
- name: "models/imagen-4.0-generate-001",
- displayName: "Imagen 4.0",
- supportedGenerationMethods: ["predict"],
- },
- {
- name: "models/text-embedding-004",
- displayName: "Text Embedding 004",
- supportedGenerationMethods: ["embedContent"],
- },
- {
- name: "models/gemini-live-2.5-flash",
- displayName: "Gemini Live",
- supportedGenerationMethods: ["bidiGenerateContent"],
- },
- {
- name: "models/veo-3.0-generate-001",
- displayName: "Veo 3.0",
- supportedGenerationMethods: ["predictLongRunning"],
- },
- {
- // Defensive: an Imagen model exposed via a long-running method must stay
- // "images", never "video".
- name: "models/imagen-future-preview",
- displayName: "Imagen Future",
- supportedGenerationMethods: ["predictLongRunning"],
- },
- ],
-};
-
-test("parseGeminiModelsList strips the models/ prefix and maps display name", () => {
- const models = parseGeminiModelsList(SAMPLE);
- const flash = models.find((m) => m.id === "gemini-2.5-flash");
- assert.ok(flash, "gemini-2.5-flash should be present");
- assert.equal(flash!.name, "Gemini 2.5 Flash");
- assert.equal(flash!.inputTokenLimit, 1048576);
- assert.equal(flash!.outputTokenLimit, 65536);
- assert.equal(flash!.supportsThinking, true);
- assert.deepEqual(flash!.supportedEndpoints, ["chat"]);
-});
-
-test("parseGeminiModelsList maps generateContent image models to the chat endpoint", () => {
- const models = parseGeminiModelsList(SAMPLE);
- const proImage = models.find((m) => m.id === "gemini-3-pro-image-preview");
- assert.ok(proImage, "gemini-3-pro-image-preview should be present");
- // gemini-*-image models generate images via generateContent (chat-with-modalities path).
- assert.deepEqual(proImage!.supportedEndpoints, ["chat"]);
-});
-
-test("parseGeminiModelsList maps Imagen predict models to the images endpoint", () => {
- const models = parseGeminiModelsList(SAMPLE);
- const imagen = models.find((m) => m.id === "imagen-4.0-generate-001");
- assert.ok(imagen, "imagen-4.0-generate-001 should be present");
- assert.deepEqual(imagen!.supportedEndpoints, ["images"]);
-});
-
-test("parseGeminiModelsList maps embedContent and bidiGenerateContent", () => {
- const models = parseGeminiModelsList(SAMPLE);
- assert.deepEqual(
- models.find((m) => m.id === "text-embedding-004")!.supportedEndpoints,
- ["embeddings"]
- );
- assert.deepEqual(
- models.find((m) => m.id === "gemini-live-2.5-flash")!.supportedEndpoints,
- ["audio"]
- );
-});
-
-test("parseGeminiModelsList maps Veo predictLongRunning models to the video endpoint", () => {
- const models = parseGeminiModelsList(SAMPLE);
- const veo = models.find((m) => m.id === "veo-3.0-generate-001");
- assert.ok(veo, "veo-3.0-generate-001 should be present");
- assert.deepEqual(veo!.supportedEndpoints, ["video"]);
-});
-
-test("parseGeminiModelsList keeps Imagen as images even via a long-running method", () => {
- const models = parseGeminiModelsList(SAMPLE);
- const imagen = models.find((m) => m.id === "imagen-future-preview");
- assert.ok(imagen, "imagen-future-preview should be present");
- assert.deepEqual(imagen!.supportedEndpoints, ["images"]);
-});
-
-test("parseGeminiModelsList defaults to chat and tolerates empty/missing input", () => {
- assert.deepEqual(parseGeminiModelsList({}), []);
- assert.deepEqual(parseGeminiModelsList(null), []);
- const [m] = parseGeminiModelsList({ models: [{ name: "models/mystery" }] });
- assert.equal(m.id, "mystery");
- assert.deepEqual(m.supportedEndpoints, ["chat"]);
-});
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import { parseGeminiModelsList } from "../../src/lib/providerModels/geminiModelsParser";
+
+// A representative slice of the live generativelanguage v1beta/models response — including the
+// image models (gemini-*-image via generateContent, imagen-* via predict) that the Vertex catalog
+// must surface dynamically.
+const SAMPLE = {
+ models: [
+ {
+ name: "models/gemini-2.5-flash",
+ displayName: "Gemini 2.5 Flash",
+ inputTokenLimit: 1048576,
+ outputTokenLimit: 65536,
+ supportedGenerationMethods: ["generateContent", "countTokens", "batchGenerateContent"],
+ thinking: true,
+ },
+ {
+ name: "models/gemini-3-pro-image-preview",
+ displayName: "Gemini 3 Pro Image Preview",
+ supportedGenerationMethods: ["generateContent", "countTokens"],
+ },
+ {
+ name: "models/imagen-4.0-generate-001",
+ displayName: "Imagen 4.0",
+ supportedGenerationMethods: ["predict"],
+ },
+ {
+ name: "models/text-embedding-004",
+ displayName: "Text Embedding 004",
+ supportedGenerationMethods: ["embedContent"],
+ },
+ {
+ name: "models/gemini-live-2.5-flash",
+ displayName: "Gemini Live",
+ supportedGenerationMethods: ["bidiGenerateContent"],
+ },
+ {
+ name: "models/veo-3.0-generate-001",
+ displayName: "Veo 3.0",
+ supportedGenerationMethods: ["predictLongRunning"],
+ },
+ {
+ // Defensive: an Imagen model exposed via a long-running method must stay
+ // "images", never "video".
+ name: "models/imagen-future-preview",
+ displayName: "Imagen Future",
+ supportedGenerationMethods: ["predictLongRunning"],
+ },
+ ],
+};
+
+test("parseGeminiModelsList strips the models/ prefix and maps display name", () => {
+ const models = parseGeminiModelsList(SAMPLE);
+ const flash = models.find((m) => m.id === "gemini-2.5-flash");
+ assert.ok(flash, "gemini-2.5-flash should be present");
+ assert.equal(flash!.name, "Gemini 2.5 Flash");
+ assert.equal(flash!.inputTokenLimit, 1048576);
+ assert.equal(flash!.outputTokenLimit, 65536);
+ assert.equal(flash!.supportsThinking, true);
+ assert.deepEqual(flash!.supportedEndpoints, ["chat"]);
+});
+
+test("parseGeminiModelsList maps generateContent image models to the chat endpoint", () => {
+ const models = parseGeminiModelsList(SAMPLE);
+ const proImage = models.find((m) => m.id === "gemini-3-pro-image-preview");
+ assert.ok(proImage, "gemini-3-pro-image-preview should be present");
+ // gemini-*-image models generate images via generateContent (chat-with-modalities path).
+ assert.deepEqual(proImage!.supportedEndpoints, ["chat"]);
+});
+
+test("parseGeminiModelsList maps Imagen predict models to the images endpoint", () => {
+ const models = parseGeminiModelsList(SAMPLE);
+ const imagen = models.find((m) => m.id === "imagen-4.0-generate-001");
+ assert.ok(imagen, "imagen-4.0-generate-001 should be present");
+ assert.deepEqual(imagen!.supportedEndpoints, ["images"]);
+});
+
+test("parseGeminiModelsList maps embedContent and bidiGenerateContent", () => {
+ const models = parseGeminiModelsList(SAMPLE);
+ assert.deepEqual(models.find((m) => m.id === "text-embedding-004")!.supportedEndpoints, [
+ "embeddings",
+ ]);
+ assert.deepEqual(models.find((m) => m.id === "gemini-live-2.5-flash")!.supportedEndpoints, [
+ "audio",
+ ]);
+});
+
+test("parseGeminiModelsList maps Veo predictLongRunning models to the video endpoint", () => {
+ const models = parseGeminiModelsList(SAMPLE);
+ const veo = models.find((m) => m.id === "veo-3.0-generate-001");
+ assert.ok(veo, "veo-3.0-generate-001 should be present");
+ assert.deepEqual(veo!.supportedEndpoints, ["video"]);
+});
+
+test("parseGeminiModelsList keeps Imagen as images even via a long-running method", () => {
+ const models = parseGeminiModelsList(SAMPLE);
+ const imagen = models.find((m) => m.id === "imagen-future-preview");
+ assert.ok(imagen, "imagen-future-preview should be present");
+ assert.deepEqual(imagen!.supportedEndpoints, ["images"]);
+});
+
+test("parseGeminiModelsList defaults to chat and tolerates empty/missing input", () => {
+ assert.deepEqual(parseGeminiModelsList({}), []);
+ assert.deepEqual(parseGeminiModelsList(null), []);
+ const [m] = parseGeminiModelsList({ models: [{ name: "models/mystery" }] });
+ assert.equal(m.id, "mystery");
+ assert.deepEqual(m.supportedEndpoints, ["chat"]);
+});
diff --git a/tests/unit/kiro-available-models.test.ts b/tests/unit/kiro-available-models.test.ts
index d1187b9608..4f57ec1a92 100644
--- a/tests/unit/kiro-available-models.test.ts
+++ b/tests/unit/kiro-available-models.test.ts
@@ -1,147 +1,163 @@
-import test from "node:test";
-import assert from "node:assert/strict";
-
-import {
- parseKiroModels,
- resolveKiroRegion,
- buildKiroModelsEndpoints,
- fetchKiroAvailableModels,
-} from "../../open-sse/services/kiroModels.ts";
-
-const FALLBACK = [{ id: "auto-kiro", name: "Auto" }, { id: "claude-sonnet-4.6" }];
-
-function jsonResponse(body: unknown, status = 200): Response {
- return new Response(JSON.stringify(body), {
- status,
- headers: { "Content-Type": "application/json" },
- });
-}
-
-test("parseKiroModels reads CodeWhisperer ListAvailableModels shape", () => {
- const models = parseKiroModels({
- models: [
- { modelId: "auto", modelName: "Auto" },
- { modelId: "claude-sonnet-4.6", modelName: "Claude Sonnet 4.6" },
- { modelId: "claude-sonnet-4.6" }, // duplicate id is ignored
- { modelName: "no id" }, // missing id is skipped
- ],
- });
-
- assert.deepEqual(
- models.map((m) => m.id),
- ["auto", "claude-sonnet-4.6"]
- );
- assert.equal(models[1].name, "Claude Sonnet 4.6");
- assert.equal(models[0].owned_by, "kiro");
-});
-
-test("resolveKiroRegion prefers stored region, then profileArn, else us-east-1", () => {
- assert.equal(resolveKiroRegion({ region: "eu-central-1" }), "eu-central-1");
- assert.equal(
- resolveKiroRegion({ profileArn: "arn:aws:codewhisperer:eu-central-1:123:profile/X" }),
- "eu-central-1"
- );
- assert.equal(resolveKiroRegion({}), "us-east-1");
- assert.equal(resolveKiroRegion(null), "us-east-1");
-});
-
-test("buildKiroModelsEndpoints is region-matched with a us-east-1 fallback", () => {
- assert.deepEqual(buildKiroModelsEndpoints("us-east-1"), [
- "https://q.us-east-1.amazonaws.com/ListAvailableModels",
- ]);
- assert.deepEqual(buildKiroModelsEndpoints("eu-central-1"), [
- "https://q.eu-central-1.amazonaws.com/ListAvailableModels",
- "https://q.us-east-1.amazonaws.com/ListAvailableModels",
- ]);
-});
-
-test("fetchKiroAvailableModels: simple (Builder ID) account, us-east-1, origin-only", async () => {
- const calls: string[] = [];
- const fetchImpl = (async (url: string) => {
- calls.push(url);
- return jsonResponse({ models: [{ modelId: "claude-sonnet-4.6" }, { modelId: "auto" }] });
- }) as unknown as typeof fetch;
-
- const result = await fetchKiroAvailableModels({
- accessToken: "tok",
- providerSpecificData: {}, // no region, no profileArn → us-east-1, origin-only
- fetchImpl,
- fallbackModels: FALLBACK,
- });
-
- assert.equal(result.source, "api");
- assert.deepEqual(result.models.map((m) => m.id).sort(), ["auto", "claude-sonnet-4.6"]);
- assert.deepEqual(calls, [
- "https://q.us-east-1.amazonaws.com/ListAvailableModels?origin=AI_EDITOR",
- ]);
-});
-
-test("fetchKiroAvailableModels: IAM Identity Center account, region-matched endpoint", async () => {
- const calls: string[] = [];
- const fetchImpl = (async (url: string) => {
- calls.push(url);
- // First (region-matched) endpoint succeeds.
- return jsonResponse({ models: [{ modelId: "claude-opus-4.8", modelName: "Opus 4.8" }] });
- }) as unknown as typeof fetch;
-
- const result = await fetchKiroAvailableModels({
- accessToken: "tok",
- providerSpecificData: { region: "eu-central-1" },
- fetchImpl,
- fallbackModels: FALLBACK,
- });
-
- assert.equal(result.source, "api");
- assert.deepEqual(result.models.map((m) => m.id), ["claude-opus-4.8"]);
- assert.equal(calls[0], "https://q.eu-central-1.amazonaws.com/ListAvailableModels?origin=AI_EDITOR");
-});
-
-test("fetchKiroAvailableModels: retries with profileArn when origin-only fails", async () => {
- const calls: string[] = [];
- const fetchImpl = (async (url: string) => {
- calls.push(url);
- if (url.includes("profileArn=")) {
- return jsonResponse({ models: [{ modelId: "claude-sonnet-4.6" }] });
- }
- return jsonResponse({ message: "forbidden" }, 403);
- }) as unknown as typeof fetch;
-
- const result = await fetchKiroAvailableModels({
- accessToken: "tok",
- providerSpecificData: {
- region: "us-east-1",
- profileArn: "arn:aws:codewhisperer:us-east-1:123:profile/ABC",
- },
- fetchImpl,
- fallbackModels: FALLBACK,
- });
-
- assert.equal(result.source, "api");
- assert.deepEqual(result.models.map((m) => m.id), ["claude-sonnet-4.6"]);
- // origin-only attempted first, then profileArn retry.
- assert.equal(calls.length, 2);
- assert.ok(calls[0].endsWith("?origin=AI_EDITOR"));
- assert.ok(calls[1].includes("profileArn=arn%3Aaws%3Acodewhisperer"));
-});
-
-test("fetchKiroAvailableModels: falls back to static catalog when no token", async () => {
- const result = await fetchKiroAvailableModels({
- accessToken: "",
- providerSpecificData: {},
- fallbackModels: FALLBACK,
- });
- assert.equal(result.source, "fallback");
- assert.deepEqual(result.models.map((m) => m.id), ["auto-kiro", "claude-sonnet-4.6"]);
-});
-
-test("fetchKiroAvailableModels: falls back when every upstream attempt fails", async () => {
- const fetchImpl = (async () => jsonResponse({ message: "expired" }, 403)) as unknown as typeof fetch;
- const result = await fetchKiroAvailableModels({
- accessToken: "stale",
- providerSpecificData: { region: "us-east-1" },
- fetchImpl,
- fallbackModels: FALLBACK,
- });
- assert.equal(result.source, "fallback");
- assert.deepEqual(result.models.map((m) => m.id), ["auto-kiro", "claude-sonnet-4.6"]);
-});
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import {
+ parseKiroModels,
+ resolveKiroRegion,
+ buildKiroModelsEndpoints,
+ fetchKiroAvailableModels,
+} from "../../open-sse/services/kiroModels.ts";
+
+const FALLBACK = [{ id: "auto-kiro", name: "Auto" }, { id: "claude-sonnet-4.6" }];
+
+function jsonResponse(body: unknown, status = 200): Response {
+ return new Response(JSON.stringify(body), {
+ status,
+ headers: { "Content-Type": "application/json" },
+ });
+}
+
+test("parseKiroModels reads CodeWhisperer ListAvailableModels shape", () => {
+ const models = parseKiroModels({
+ models: [
+ { modelId: "auto", modelName: "Auto" },
+ { modelId: "claude-sonnet-4.6", modelName: "Claude Sonnet 4.6" },
+ { modelId: "claude-sonnet-4.6" }, // duplicate id is ignored
+ { modelName: "no id" }, // missing id is skipped
+ ],
+ });
+
+ assert.deepEqual(
+ models.map((m) => m.id),
+ ["auto", "claude-sonnet-4.6"]
+ );
+ assert.equal(models[1].name, "Claude Sonnet 4.6");
+ assert.equal(models[0].owned_by, "kiro");
+});
+
+test("resolveKiroRegion prefers stored region, then profileArn, else us-east-1", () => {
+ assert.equal(resolveKiroRegion({ region: "eu-central-1" }), "eu-central-1");
+ assert.equal(
+ resolveKiroRegion({ profileArn: "arn:aws:codewhisperer:eu-central-1:123:profile/X" }),
+ "eu-central-1"
+ );
+ assert.equal(resolveKiroRegion({}), "us-east-1");
+ assert.equal(resolveKiroRegion(null), "us-east-1");
+});
+
+test("buildKiroModelsEndpoints is region-matched with a us-east-1 fallback", () => {
+ assert.deepEqual(buildKiroModelsEndpoints("us-east-1"), [
+ "https://q.us-east-1.amazonaws.com/ListAvailableModels",
+ ]);
+ assert.deepEqual(buildKiroModelsEndpoints("eu-central-1"), [
+ "https://q.eu-central-1.amazonaws.com/ListAvailableModels",
+ "https://q.us-east-1.amazonaws.com/ListAvailableModels",
+ ]);
+});
+
+test("fetchKiroAvailableModels: simple (Builder ID) account, us-east-1, origin-only", async () => {
+ const calls: string[] = [];
+ const fetchImpl = (async (url: string) => {
+ calls.push(url);
+ return jsonResponse({ models: [{ modelId: "claude-sonnet-4.6" }, { modelId: "auto" }] });
+ }) as unknown as typeof fetch;
+
+ const result = await fetchKiroAvailableModels({
+ accessToken: "tok",
+ providerSpecificData: {}, // no region, no profileArn → us-east-1, origin-only
+ fetchImpl,
+ fallbackModels: FALLBACK,
+ });
+
+ assert.equal(result.source, "api");
+ assert.deepEqual(result.models.map((m) => m.id).sort(), ["auto", "claude-sonnet-4.6"]);
+ assert.deepEqual(calls, [
+ "https://q.us-east-1.amazonaws.com/ListAvailableModels?origin=AI_EDITOR",
+ ]);
+});
+
+test("fetchKiroAvailableModels: IAM Identity Center account, region-matched endpoint", async () => {
+ const calls: string[] = [];
+ const fetchImpl = (async (url: string) => {
+ calls.push(url);
+ // First (region-matched) endpoint succeeds.
+ return jsonResponse({ models: [{ modelId: "claude-opus-4.8", modelName: "Opus 4.8" }] });
+ }) as unknown as typeof fetch;
+
+ const result = await fetchKiroAvailableModels({
+ accessToken: "tok",
+ providerSpecificData: { region: "eu-central-1" },
+ fetchImpl,
+ fallbackModels: FALLBACK,
+ });
+
+ assert.equal(result.source, "api");
+ assert.deepEqual(
+ result.models.map((m) => m.id),
+ ["claude-opus-4.8"]
+ );
+ assert.equal(
+ calls[0],
+ "https://q.eu-central-1.amazonaws.com/ListAvailableModels?origin=AI_EDITOR"
+ );
+});
+
+test("fetchKiroAvailableModels: retries with profileArn when origin-only fails", async () => {
+ const calls: string[] = [];
+ const fetchImpl = (async (url: string) => {
+ calls.push(url);
+ if (url.includes("profileArn=")) {
+ return jsonResponse({ models: [{ modelId: "claude-sonnet-4.6" }] });
+ }
+ return jsonResponse({ message: "forbidden" }, 403);
+ }) as unknown as typeof fetch;
+
+ const result = await fetchKiroAvailableModels({
+ accessToken: "tok",
+ providerSpecificData: {
+ region: "us-east-1",
+ profileArn: "arn:aws:codewhisperer:us-east-1:123:profile/ABC",
+ },
+ fetchImpl,
+ fallbackModels: FALLBACK,
+ });
+
+ assert.equal(result.source, "api");
+ assert.deepEqual(
+ result.models.map((m) => m.id),
+ ["claude-sonnet-4.6"]
+ );
+ // origin-only attempted first, then profileArn retry.
+ assert.equal(calls.length, 2);
+ assert.ok(calls[0].endsWith("?origin=AI_EDITOR"));
+ assert.ok(calls[1].includes("profileArn=arn%3Aaws%3Acodewhisperer"));
+});
+
+test("fetchKiroAvailableModels: falls back to static catalog when no token", async () => {
+ const result = await fetchKiroAvailableModels({
+ accessToken: "",
+ providerSpecificData: {},
+ fallbackModels: FALLBACK,
+ });
+ assert.equal(result.source, "fallback");
+ assert.deepEqual(
+ result.models.map((m) => m.id),
+ ["auto-kiro", "claude-sonnet-4.6"]
+ );
+});
+
+test("fetchKiroAvailableModels: falls back when every upstream attempt fails", async () => {
+ const fetchImpl = (async () =>
+ jsonResponse({ message: "expired" }, 403)) as unknown as typeof fetch;
+ const result = await fetchKiroAvailableModels({
+ accessToken: "stale",
+ providerSpecificData: { region: "us-east-1" },
+ fetchImpl,
+ fallbackModels: FALLBACK,
+ });
+ assert.equal(result.source, "fallback");
+ assert.deepEqual(
+ result.models.map((m) => m.id),
+ ["auto-kiro", "claude-sonnet-4.6"]
+ );
+});