Capability Map¶
This page checks the first-class capabilities currently owned by ChatVoice, their verification state, and current boundaries.
Current capabilities¶
-
Packaged web service
Installing
ChatVoice[web]lets operators start the current Speakr FastAPI + browser service withchatvoice serve app. -
Fresh-start accounts and API tokens
chatvoice accounts addcreates invited accounts in the packaged runtime database. Signed-in users can create one-time-visible API tokens from the web settings panel; the CLI can also create/list/revoke token metadata. -
Data read API / CLI
GET /api/data/...andchatvoice data ...use bearer tokens to read meeting transcripts, meeting summaries, and realtime conversation text records. -
API-first ASR provider
The production direction is
api-server: the backend calls a managed ASR API or self-hosted GPU ASR server over HTTP instead of embedding GPU runtime in the web process. -
Runtime paths and service plan
chatvoice pathsandchatvoice service planread back data, log, run, temp, and model-cache paths under ChatArch home. -
Health checks
chatvoice health statusreads/api/statusfrom a running service. -
Local one-shot voice cloning
Signed-in users can upload or record authorized reference audio in Voice Studio, enter new text, and submit a job through ChatVoice to the hitk VoiceClone sidecar / IndexTTS-2.5. The page shows progress and returns a current-job preview/download audio.
Status table¶
| Capability | Status | Notes |
|---|---|---|
| Base CLI entries | Implemented | --help, --version, and shared ChatStyle --tree / --tree-brief. |
| Runtime paths | Implemented | Default <chatarch-home>/chatvoice/, override with runtime-home overrides. |
| Packaged web startup | Implemented | chatvoice serve app calls chatvoice.web.server:create_app. |
| Invited account CLI | Implemented | chatvoice accounts add/list; passwords are read from environment variables only. |
| API token management | Implemented | Web settings panel + CLI token lifecycle; the server stores hashes only. |
| Data read API/CLI | Implemented | Bearer token reads for meetings, summaries, and realtime conversations. |
| Local one-shot voice cloning | Implemented | /api/voice-clone/* proxies the VoiceClone sidecar; no voice profile or generated-audio history is saved. |
| ASR API provider | Implemented | CHATVOICE_ASR_CHANNEL=api-server + the ASR API URL setting. |
| Local contract smoke | Implemented | CHATVOICE_ASR_CHANNEL=stub-local starts the full path without GPU/cloud credentials. |
| Local FunASR compatibility | Preserved | funasr-gpu / funasr-cpu remain available, but production should prefer an external ASR API server. |
| SQLite WAL storage | Implemented | Default for one service process and light concurrency; the api_tokens table stores only hash/prefix/metadata. |
| Postgres/MySQL storage | Not implemented | No DATABASE_URL switch is provided; future high-concurrency Postgres/MySQL support is a separate storage-layer migration. |
Out of scope now¶
- Do not bundle GPU model download, CUDA/PyTorch installation, and the web process as one default runtime.
- Do not claim MySQL/Postgres is complete in v0.1.13; high-concurrency storage migration needs a separate release.
- Do not print tokens, cookies, Authorization headers, or raw recordings; full transcripts are returned only by explicit data-read commands.
- Do not present one-shot voice cloning as a permanent voice profile; the current flow needs reference audio and target text for each generation.
- Do not manage services with
kill/kill -9; restart commands need supervisor/graceful boundaries first.