Skip to content

ChatVoice Docs

ChatVoice is a ChatArch Python package that packages the Speakr recording, transcription, meeting-notes, and voice-workspace service into an installable, maintainable, API-readable runtime.

Documentation entry: https://arch.gh.wzhecnu.cn/ChatVoice/en/

Choose by scenario

Scenario Document
Install from PyPI and start the service Deployment and Startup
Inspect install location, runtime directories, and SQLite/IndexedDB data structure Runtime Layout and Data Structure
Understand why original recordings are not saved Recording Storage Boundary
Use one-shot local voice cloning in Voice Studio Voice Cloning Guide
Generate API tokens and read meeting/summary data API Access
Read back the real command tree and boundaries CLI Tree
Check first-class capabilities and current boundaries Capability Map
Call package behavior directly from Python Python Interface Tree

Core entries

  • Deployment and Startup

    From python -m pip install "ChatVoice[web]==0.1.13" to chatvoice serve app, including runtime paths, account provisioning, ASR API provider wiring, and database concurrency boundaries.

    Read deployment guide

  • API Access

    Browser login, API token lifecycle, and chatvoice data ... reads for meeting transcripts, meeting summaries, and realtime conversation records.

    Read API access

  • Runtime Layout and Data Structure

    site-packages install path, ~/.chatarch/chatvoice runtime root, SQLite tables, IndexedDB, temp/asr, and model-cache.

    Read runtime layout

  • Recording Storage Boundary

    The meeting recorder saves text and summaries only. It does not save original recordings; future pure-recording support should be designed as a separate capability.

    Read recording storage boundary

  • Voice Cloning Guide

    Complete workflow for reference-audio upload/recording, new text entry, progress, preview playback, download, and current one-shot boundaries.

    Read voice cloning guide

  • CLI Tree

    The real implemented command tree, command status, and update rules.

    Read CLI tree

  • Capability Map

    Review current package boundaries and avoid presenting planned work as implemented functionality.

    Read capability map

  • Python Interface Tree

    Keep the CLI thin and put substantive behavior in importable Python APIs.

    Read interface tree

v0.1.13 deployment boundary

  • The packaged FastAPI app starts with chatvoice serve app.
  • Fresh start can create invited accounts with chatvoice accounts add; no source-tree script is required.
  • Signed-in users can create API tokens in the web UI; the CLI can use tokens to read meetings, summaries, and realtime conversations.
  • Signed-in users can upload or record authorized reference audio in Voice Studio and run a one-shot VoiceClone/IndexTTS-2.5 preview through the hitk sidecar.
  • The meeting recorder does not save or download original recordings; the server stores text, summaries, and metadata only.
  • Production ASR should use api-server against a managed API or self-hosted GPU ASR server.
  • stub-local is only for credential-free / GPU-free contract smoke.
  • v0.1.13 defaults to SQLite WAL for one service process and light concurrency; high-concurrency storage migration needs a separate release.

Preview docs locally

python -m pip install -e ".[docs]"
mkdocs serve

Chinese home is available at https://arch.gh.wzhecnu.cn/ChatVoice/.