Add PostHog product events for manual TTS stop clicks and conversation controls across Web, mobile, and Electron chat surfaces. Document the events in metrics ownership and cover the analytics API with Vitest.
Signed-off-by: RainbowBird <git@luoling.moe>
Commit-Message-Assisted-by: Codex
Track per-app TTS concurrency in Redis, route capped upstreams by
available pool capacity, and surface pool saturation metrics. Document
the Voice Pack plan so the remaining backend and card-binding work has
an explicit implementation map.
Upgrade @moeru/eventa to v1.0.0-beta.8 through the pnpm catalog and remove the local Hono adapter copy.
Keep AIRI on business-level Eventa usage while adapter contract tests live upstream in @moeru/eventa.
Signed-off-by: RainbowBird <git@luoling.moe>
Commit-Message-Assisted-by: Claude (via Claude Code)
Separate OpenAI-compatible route wiring from chat, speech, catalog, billing, and telemetry pipeline code.
Signed-off-by: RainbowBird <git@luoling.moe>
Commit-Message-Assisted-by: Claude (via Claude Code)
The packaged Electron renderer loads from file://, so its Stripe checkout
and portal requests carry no Referer and an opaque/absent Origin.
resolveTrustedRequestOrigin() returned undefined and the routes threw
INVALID_ORIGIN ("Missing trusted request origin"), blocking FLUX purchases
on desktop. Web and mobile were unaffected because they send a trusted web
origin. CORS was not the blocker: file:// requests already reach the API
(balance, providers, characters all work), so only the checkout-specific
redirect-base dependency failed.
Stripe success_url/cancel_url/return_url must be http(s) URLs, so file://
can never be a valid redirect base. Replace the trusted-origin requirement
with resolveCheckoutRedirectBase(), which prefers the request's trusted
origin (web/mobile return to where they started) and falls back to a
configured canonical web app URL (WEB_APP_URL, default https://airi.moeru.ai)
for origin-less clients like the desktop app.
On the client, the desktop app opens checkout in the system browser via
window.open (routed through setWindowOpenHandler -> shell.openExternal)
instead of navigating the file:// window to Stripe, and refreshes the FLUX
balance on window focus since it never receives the success_url redirect.
The streaming speech provider settings page was seeding its model
picker with a hardcoded `volcengine/seed-tts-2.0`, which contradicts
the just-landed "streaming model catalog comes from the server"
contract. Now the default also comes from configKV.
- `UNSPEECH_UPSTREAM.streaming` gains `defaultModel?: string`.
- `GET /api/v1/audio/models/streaming` response gains `default: string
| null` reading that field.
- admin slice apply: `streaming.defaultModel` survives key/URL
rotation alongside `streaming.models`.
- `providerOfficialSpeechStreaming.listModels` stashes the value in a
module-scope ref, exposed via `getDefaultStreamingModel()`.
- Streaming speech settings page reads the helper instead of the
hardcoded id; falls back to the first server-returned model if the
operator hasn't curated a default. `handleGenerateSpeech` no longer
has a `seed-tts-2.0` backstop — a missing backend prefix in the
selected model id now throws instead of silently defaulting.
One unspeech deployment = one configKV entry. The split into two keys
modelled a hypothetical split deployment (operators pointing REST and
WS at different unspeech instances) that nobody actually runs, and the
explanatory comments justifying the split were exactly the migration-
narrative anti-pattern we just banned from source.
Schema:
UNSPEECH_UPSTREAM = {
restBaseURL: string, // required, http(s)://host:port
streaming?: { // optional, only when ws TTS is on
baseURL: string, // ws(s)://host:port/...
keys: [{ id, ciphertext }],
adapterParams: {},
models?: [{ id, name?, description? }],
},
}
Admin slice surface flattens to one `kind: 'unspeech'`:
{ kind: 'unspeech', restBaseURL,
streaming?: { upstreamURL, plaintextKey, keyEntryId? } }
Read-site changes:
- routeTts + listTtsVoices read UNSPEECH_UPSTREAM.restBaseURL via
getOrThrow; absent entry → 503 CONFIG_NOT_SET.
- audio-speech-ws dials UNSPEECH_UPSTREAM.streaming and 1008-closes
with streaming_tts_not_configured when the subtree is absent.
- handleListStreamingVoices reads .restBaseURL directly (no more
ws→http scheme swap) and 503s on missing streaming subtree.
- handleListStreamingTTSModels reads .streaming.models.
- config-sync-subscriber listens for UNSPEECH_UPSTREAM invalidations
instead of UNSPEECH_REST_BASE_URL.
Drive-by: tighten ConfigKVService.getOrThrow return to
`Exclude<ConfigDefinitions[K], undefined>` so call sites stop needing
non-null assertions on optional schema entries — the runtime already
throws, the type now reflects it.
Migration: operators repost a single admin slice to rewrite the merged
entry. Old STREAMING_TTS_UPSTREAM and UNSPEECH_REST_BASE_URL rows
become dead data and can be deleted from configKV; nothing reads them.
Frontend voice picker silently lost Azure voices because /audio/voices
was called without ?model=, so the server fell back to DEFAULT_TTS_MODEL
and only ever returned one model's catalog. /audio/models also hardcoded
a single `auto` alias, so the client could never request a specific
model id even if it wanted to. On top of that, the Azure adapter shipped
only 2 hand-typed voices in voices/azure.json — the rest of Microsoft's
catalog was unreachable.
Fixed in one pass:
- Drop voices/*.json. Every getVoiceCatalog now calls unspeech REST
(backend=microsoft|alibaba|volcengine). Azure proxies Microsoft's live
voices/list; cosyvoice + volcengine use unspeech's embedded catalogs.
- Drop direct upstream provider calls from send(). All three adapters
POST to <UNSPEECH_REST_BASE_URL>/v1/audio/speech with the openai-shaped
body unspeech expects (model: `<backend>/<id>`). Azure SSML still built
on our side so <prosody rate> speed survives — unspeech accepts
pre-built <speak> documents.
- Fail-fast on every voices endpoint: empty ?model= → 400 MISSING_MODEL
(no implicit DEFAULT_TTS_MODEL fallback), only `auto` resolves;
streaming upstream missing → 503; unspeech non-2xx / network err → 502;
malformed body → 502 with grepable message. No silent `{voices: []}`.
- /audio/models lists every LLM_ROUTER_CONFIG.tts.models key + `auto`.
- Frontend providerOfficialSpeech.listVoices passes ?model= and throws
on upstream failure instead of returning [].
- New UNSPEECH_REST_BASE_URL configKV entry (no default, missing → 503).
config-sync subscriber invalidates the voice cache when this key OR
LLM_ROUTER_CONFIG changes.
- Router voice catalog Redis cache covers all providers with per-provider
TTL (azure 6h live, alibaba/volcengine 24h since unspeech embeds them
at build time).
Pending billing-SKU verification: DashScope cosyvoice now goes through
unspeech's WS-internal alibaba backend (was REST `/SpeechSynthesizer`).
Functional output equivalent — confirm bill after deploy, revert via
git history if WS pricing is materially higher.
Streaming TTS model list was hardcoded in the official provider with two
Volcengine ids. This moves it server-side so operators control the
catalog without a client release.
Catalog lives on `STREAMING_TTS_UPSTREAM.models` (same configKV entry as
the ws upstream + key envelope) rather than a separate kv key — connection
target, keys, and surfaced ids are one deployment decision, splitting
them risks drift on rotation. `LLM_ROUTER_CONFIG.tts.models` is the
router/fallback domain for HTTP TTS; streaming is single-ws and has no
router semantics, so it gets its own `streamingTtsUpstreamSchema`
instead of reusing `ttsUpstreamSchema`.
New `GET /api/v1/audio/models/streaming` returns the configured list;
empty when upstream or models is unset (UI renders "no models" instead
of 5xx). Admin slice apply preserves existing `models` across key/connection
rotation so admin POSTs that only carry upstream+keys do not wipe the
catalog. Frontend `providerOfficialSpeechStreaming.listModels` fetches
the endpoint and throws on upstream errors (no silent empty array).
- services/domain/llm-router and services/domain/user-deletion still had
colocated *.test.ts; moved them into the local tests/ subdir so every
service module follows the same layout as billing/, admin/*/, etc.
- routes/oidc/ wasn't an independent route group — it's wholly consumed
by routes/auth/index.ts (electron-callback relay + OIDC bearer token
helper). Moved to routes/auth/oidc/ to reflect that ownership.
- Fixed the relative imports in src/libs/tests/{auth,env,request-auth,
ws-auth}.test.ts. They were moved to libs/tests/ in the previous
commit without updating the './<source>' specifiers; typecheck only
caught it once the rewriter ran across the tree.
pnpm typecheck passes; eslint matches the main baseline.
Why
- Code, routes, service, worker, tests, and ai-context references for the
legacy flux_grant_batch flow were removed in the previous commit. The
schema file and the corresponding production tables were intentionally
left for a separate DDL-only PR (this one) so the destructive change is
easy to time and roll back.
What
- Delete src/schemas/flux-grant-batch.ts.
- Drop the re-export line in src/schemas/index.ts.
- drizzle-kit generate produced drizzle/0011_open_unus.sql:
DROP TABLE flux_grant_batch CASCADE;
DROP TABLE flux_grant_batch_recipient CASCADE;
CASCADE removes the 6 associated indexes in one shot.
- docs/ai-context/architecture-overview.md updated: the dead-code
reminder now points at the migration and explains the rollback story.
Also rolls in a pre-existing local move that was sitting uncommitted:
src/libs/{auth,env,request-auth,ws-auth}.test.ts → src/libs/tests/...
(aligning with the libs/tests/eventa-hono-adapter.test.ts placement that
was already on HEAD).
Deployment
- pnpm typecheck: passes.
- DO NOT run pnpm db:push on prod from this branch automatically. The
drop is intentionally a separate operator action that requires picking
a deploy window where no instance is still on an older image that
could try to read flux_grant_batch. Until 0011 is applied to prod the
table sits as an orphaned shell — safe to leave indefinitely.
Why
- src/services/ was an unordered mix of single-file services and module
directories with no shared classification axis, plus several long-dead
admin batch helpers that survived the move to the simpler synchronous
admin-flux-grants flow.
What
- services/ now has two top-level layers:
domain/ — DB state + business rules (billing, characters, chats,
flux, flux-transaction, llm-router, providers, request-log,
stripe, user-deletion, admin/{flux-grants,router-config})
adapters/ — thin wrappers over external SDKs / infra (config-kv, email,
posthog, tts/)
- admin/* moved under domain/admin/ with consistent plural names
(flux-grants, router-config).
- tts-adapters/ collapsed to adapters/tts/ (no redundant -adapters suffix
once nested under adapters/).
- 63 src files + scripts/e2e-llm-router.ts + tests/verifications/_harness.ts
had relative imports rewritten; git mv preserves blame.
- apps/server/CLAUDE.md and docs/ai-context/*.md updated to match new paths.
Dead code removed
- services/admin-flux-grant-batches/ (service + worker + tests, 1090 LOC) —
superseded by admin-flux-grants and never wired into app.ts.
- routes/admin/flux-grant-batches/ — same.
- utils/redis-compressed.ts + test — zero production call sites.
- llm-router/index.ts re-exports trimmed from 26 to 6; only symbols with
external consumers are kept.
Intentionally kept
- schemas/flux-grant-batch.ts and its schemas/index.ts export remain so the
drizzle-kit generate diff stays empty. Removing them is a separate PR
that owns the drop-table migration for flux_grant_batch /
flux_grant_batch_recipient.
Verification
- pnpm -F @proj-airi/server typecheck: passes.
- pnpm exec eslint apps/server: 49 errors, identical to main baseline
(all are pre-existing node/prefer-global/buffer in envelope-crypto and
scripts/e2e-llm-router; untouched by this change).
- Vitest passes per-file; the 6 mockDB hook timeouts under full-parallel
run are the known pushSchema-per-worker infra cost, not a regression.
The two seed scripts are fully superseded by the new admin endpoint
`POST /api/admin/config/router` — same encryption, same configKV
writes, same `configkv:invalidate` publish, plus auth + audit + body
limits. Keeping both code paths created a drift risk on the AAD label
and the merge semantics.
Doc + test fallout:
- `e2e-llm-router.ts` now points readers to the admin endpoint for
the prerequisite seed step.
- `docs/ai-context/verifications/llm-router.md` and
`streaming-tts.md` get curl-based seed instructions; the 2026-05-15
llm-router evidence stays intact with a note that the script it
used has since been removed.
- The U9 follow-up entry in `llm-router.md` flips from "not shipped"
to "partially shipped" — ETag + HMAC publish are still deferred,
so the `config_write` / `config_invalid_hmac` Grafana panels stay
parked.
- Self-edit on the admin route + `app.ts` docstrings to drop the
earlier "scripts stay as break-glass" wording.
RainbowBird