Make your computer a home for your AI partner. Avibe is the local-first Agent OS: Vibey, its built-in AI partner, works out of the box, and Claude Code, Codex & OpenCode run unchanged beside it — reach them from your browser, phone, or any chat app.
Python agentagent-osaiai-agentsai-partnerchatopsclaudeclaude-codecodexdevtoolsdiscord-botlark-botllmlocal-firstlong-term-memoryopencodeslack-botvibe-codingvibeywechat
import type { TranslationKey } from '@/i18n/types';
// Provenance click-through for a harness trigger message in Chat (contract
// A9a/A9b). A pure mapper so the branching is unit-tested without the component.
//
// - A9a "自动触发" (agent callback): link to the SOURCE session's chat, labelled
// by its title prefix (fallback: source agent name + short session id).
// - A9b "定时任务" / "Watch 监听": link to the matching Harness tab filtered to
// this session, reusing the backgroundActivity deep-link helper.
// Other harness rows (webhook, or an agent callback whose source didn't resolve)
// stay non-navigating.
import type { WorkbenchMessage } from '../context/ApiContext';
import { harnessNavPath } from './backgroundActivity';
export type ChatTriggerLink =
| { kind: 'source'; to: string; label: string }
| { kind: 'harness'; to: string };
const AGENT_RUN_NATIVE_PREFIX = 'agent_run:';
const HARNESS_NATIVE_PREFIXES = ['agent_run:', 'scheduled:', 'watch:', 'webhook:', 'hook:'];
// Live Harness rows can arrive before read-side enrichment. Agent callbacks need
// source-session resolution; legacy queued triggers may need kind/definition
// recovery from their stable native id. A targeted reconcile fills either set.
export function needsHarnessProvenanceReconcile(
message: Pick<
WorkbenchMessage,
'source' | 'native_message_id' | 'source_session_id' | 'author_name'
>,
): boolean {
if (message.source !== 'harness' || typeof message.native_message_id !== 'string') return false;
if (message.author_name == null) {
return HARNESS_NATIVE_PREFIXES.some((prefix) => message.native_message_id?.startsWith(prefix));
}
return (
message.native_message_id.startsWith(AGENT_RUN_NATIVE_PREFIX) &&
message.source_session_id == null
);
}
// author_name values that map to the Harness "tasks" tab (watch → "watches").
// Includes the legacy/queued-restore `task` trigger kind alongside scheduled/task_run.
const TASK_KINDS = new Set(['scheduled', 'task_run', 'task']);
const TITLE_PREFIX_MAX = 12;
function titlePrefix(title: string): string {
const trimmed = title.trim();
return trimmed.length > TITLE_PREFIX_MAX ? `${trimmed.slice(0, TITLE_PREFIX_MAX)}…` : trimmed;
}
type TriggerFields = Pick<
WorkbenchMessage,
| 'source'
| 'author_name'
| 'author_id'
| 'session_id'
| 'source_session_id'
| 'source_session_title'
| 'source_session_agent_name'
> & { metadata?: Record<string, unknown> };
const VAULT_CALLBACK_STATUS_KEYS: Record<string, TranslationKey> = {
'delete:approved': 'chat.source.vaultDeleteApproved',
'delete:fulfilled': 'chat.source.vaultDeleteApproved',
'delete:denied': 'chat.source.vaultDeleteDenied',
'delete:expired': 'chat.source.vaultDeleteExpired',
'provision:fulfilled': 'chat.source.vaultProvided',
'access:approved': 'chat.source.vaultAccessApproved',
'sign:approved': 'chat.source.vaultSigned',
'access:denied': 'chat.source.vaultDenied',
'sign:denied': 'chat.source.vaultDenied',
'provision:denied': 'chat.source.vaultDenied',
'access:failed': 'chat.source.vaultFailed',
'sign:failed': 'chat.source.vaultFailed',
'provision:failed': 'chat.source.vaultFailed',
'access:expired': 'chat.source.vaultExpired',
'sign:expired': 'chat.source.vaultExpired',
'provision:expired': 'chat.source.vaultExpired',
};
function vaultCallbackMetadata(message: TriggerFields): { requestType: string; status: string } | null {
if (message.source !== 'harness') return null;
const metadata = message.metadata;
const sourceActor = typeof metadata?.source_actor === 'string' ? metadata.source_actor : '';
if (!sourceActor.startsWith('vault:')) return null;
const requestType = typeof metadata?.vault_request_type === 'string' ? metadata.vault_request_type : '';
const status = typeof metadata?.vault_request_status === 'string' ? metadata.vault_request_status : '';
return { requestType, status };
}
export function isVaultCallback(message: TriggerFields): boolean {
if (message.source !== 'harness') return false;
const metadata = message.metadata;
const sourceActor = typeof metadata?.source_actor === 'string' ? metadata.source_actor : '';
return sourceActor.startsWith('vault:');
}
export function vaultCallbackStatusKey(message: TriggerFields): TranslationKey | null {
const metadata = vaultCallbackMetadata(message);
if (!metadata) return null;
return VAULT_CALLBACK_STATUS_KEYS[`${metadata.requestType}:${metadata.status}`] ?? null;
}
// ``agentFallback`` is the localized word for "agent" (chat.source.agentFallback);
// passed in so this stays a pure, translation-free mapper.
export function chatTriggerLink(message: TriggerFields, agentFallback: string): ChatTriggerLink | null {
if (message.source !== 'harness') return null;
const sourceId = message.source_session_id;
if (sourceId) {
const title = message.source_session_title?.trim();
const label = title
? titlePrefix(title)
: `${message.source_session_agent_name?.trim() || agentFallback} · ${sourceId.slice(-6)}`;
return { kind: 'source', to: `/chat/${encodeURIComponent(sourceId)}`, label };
}
const kind = message.author_name;
// ``show_intent`` (a page button action) stays a non-navigating harness row.
// ``show_annotation`` no longer reaches here at all: an annotation is typed
// ``annotation`` and renders as its own card, never as a harness trigger row.
if (kind === 'show_intent') return null;
if (kind === 'watch' || TASK_KINDS.has(kind ?? '')) {
const itemKind = kind === 'watch' ? 'watch' : 'task';
return {
kind: 'harness',
to: harnessNavPath({ id: `${itemKind}:${message.author_id ?? ''}`, item_kind: itemKind }, message.session_id),
};
}
return null;
}
// i18n key for a harness chip's leading label. Branches on source presence:
// - a resolved agent callback (source_session_id set) uses the "From" PREFIX
// (chat.source.from) that leads into the source-session link → "From · <title>↗".
// - an unresolved one (source deleted, or a pre-enrichment live row) falls back to
// the SELF-CONTAINED "From agent" (chat.source.harness) — still source semantics,
// never the mechanism-flavored "Automated" that read as system-triggered.
// - Task / Watch / Webhook keep their own self-contained labels.
// Pure so the branch is unit-tested alongside chatTriggerLink.
export function harnessChipLabelKey(message: TriggerFields): TranslationKey {
if (message.source === 'harness' && message.source_session_id) return 'chat.source.from';
if (isVaultCallback(message)) return 'chat.source.vault';
const kind = message.author_name;
if (kind === 'show_intent') return 'chat.source.showIntent';
if (kind === 'watch') return 'chat.source.watch';
if (kind === 'webhook') return 'chat.source.webhook';
if (kind === 'hook') return 'chat.source.hook';
if (kind === 'activity_recovery') return 'chat.source.activityRecovery';
if (TASK_KINDS.has(kind ?? '')) return 'chat.source.scheduled';
return 'chat.source.harness';
}