Laurence
51321aa7c5
Lets the owner point transcription and summarisation at any AI: an open-standard
endpoint (OpenAI-compatible / local faster-whisper / Ollama) or a commercial API
(OpenAI, Anthropic, Gemini). Config-driven, self-hostable, no lock-in.
What changed:
- server/app/providers/: provider layer.
- base.py: Transcriber/Summariser protocols + shared summary prompt + tolerant JSON
parser (uniform Summary shape across providers).
- summary.py: OpenAICompatibleSummariser (any /chat/completions - OpenAI, Groq,
OpenRouter, LocalAI, LM Studio, vLLM, Ollama /v1) and AnthropicSummariser (Claude
via the official anthropic SDK; Messages API has no OpenAI-compatible endpoint).
- transcription.py: OpenAICompatibleTranscriber (/audio/transcriptions - OpenAI,
Groq, self-hosted whisper server) and LocalWhisperTranscriber (faster-whisper,
execution wired in M5).
- factory.py: builds the configured providers with per-provider defaults
(anthropic -> claude-opus-4-8, openai_compatible -> gpt-4o-mini, ollama -> llama3.1).
- config.py + .env.example: transcription_provider / llm_provider selectors + base_url,
key, model settings; local faster-whisper and Ollama kept as the self-hosted defaults.
- main.py: /health now reports the resolved provider names (no secrets).
- requirements.txt: httpx drives all HTTP providers; anthropic + faster-whisper are
optional, only for their respective providers.
- docs/ai-providers.md: config recipes for OpenAI, Groq, Anthropic, Gemini, LocalAI,
LM Studio, Ollama, self-hosted whisper.
- state/: DECISIONS, ARCHITECTURE, TODO updated.
Why:
- The user asked to connect the device to any open standard AI or commercial one; this
is also the core differentiator vs Plaud's locked cloud.
Notes:
- Anthropic provider uses the official SDK and defaults to claude-opus-4-8 (per the
claude-api guidance). AI deps are optional per chosen provider. Modules byte-compile
cleanly; end-to-end wiring into the ingest pipeline lands with M5.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
|
||
|---|---|---|
| .forgejo/workflows | ||
| api | ||
| app | ||
| case | ||
| docs | ||
| firmware | ||
| hardware | ||
| LICENSES | ||
| server | ||
| state | ||
| .gitignore | ||
| CLAUDE.md | ||
| LICENSE | ||
| LICENSING.md | ||
| README.md | ||
OpenScribe
An open-source, self-hosted AI voice recorder in the style of the Plaud Note / NotePin. A small wearable device records audio to local storage, syncs to a mobile app and to a server you run, and produces transcripts and summaries, through a completely open API. No proprietary cloud, no lock-in.
Status: early scaffold (M0). The design, hardware BOM, and open API are defined; component code is being built milestone by milestone. See
state/TODO.md.
Why
Commercial AI recorders send your conversations to a vendor's cloud. OpenScribe keeps the whole pipeline (capture, storage, transcription, summarisation, export) on hardware you own, under open licences, with an API documented so you can build against it.
What it does
- Records voice to microSD on a small ESP32-S3 device (button start/stop, LED status).
- Syncs three ways:
- BLE for control, status and WiFi provisioning.
- WiFi to the mobile app for fast bulk transfer.
- Independent WiFi upload to your storage when on charge / hard-powered, with no phone present.
- Transcribes (faster-whisper) and summarises (local LLM via Ollama) on a server you host.
- Exposes a completely open REST API with exports: audio, TXT, SRT, VTT, Markdown, JSON.
- Ships a Flutter app (Android + iOS) and a 3D-printed parametric case.
Architecture at a glance
mic -> ESP32-S3 (record to microSD)
| BLE: control / provision
| WiFi REST API: transfer to app
| WiFi uploader (on charge) -> object storage (MinIO / WebDAV / NAS)
|
self-hosted server (FastAPI)
faster-whisper -> Ollama -> open API + exports
|
Flutter app (Android + iOS)
Full detail: state/ARCHITECTURE.md. The API contract:
api/openapi.yaml.
Repository layout
firmware/ ESP32-S3 recorder firmware (PlatformIO, Arduino-ESP32)
server/ Self-hosted FastAPI server: ingest, transcribe, summarise, serve API
app/ Flutter mobile app (Android + iOS)
hardware/ BOM, wiring/pinout, build notes (off-the-shelf modules, no PCB in v1)
case/ Parametric 3D-printed enclosure (OpenSCAD)
api/ openapi.yaml - the single source of truth for the open API
docs/ Workflow docs (this project follows the Default Workflow SOP)
state/ Project memory: PROJECT, ARCHITECTURE, DECISIONS, TODO, NOTES
LICENSES/ Full licence texts
Hardware
You build the device from off-the-shelf parts. See hardware/BOM.md
for the shopping list, wiring and pinout. Core: an ESP32-S3 board with PSRAM, an I2S MEMS
microphone, a microSD card, a LiPo battery and a charge IC.
Self-hosted stack
Everything runs on hardware you own: Forgejo for the forge, faster-whisper for transcription, Ollama for summaries, MinIO (or WebDAV / NAS) for storage. Plain recording and transfer work without any server; the server adds transcription and summaries.
Getting started
Nothing to run end to end yet (M0 scaffold). To follow along:
- Read
state/PROJECT.mdandstate/TODO.md. - Per-component build/run commands are in
state/NOTES.mdand each component'sREADME.md.
Licensing
Multi-part, copyleft. Code is GPL-3.0-only, hardware is CERN-OHL-S-2.0, case and docs are
CC-BY-SA-4.0. See LICENSING.md.
OpenScribe is an independent open-source project inspired by the Plaud class of device. It is not affiliated with or endorsed by Plaud.