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>
Bootstrap of the project (M0). Sets up the monorepo, design docs, hardware BOM,
the open API contract, component skeletons, licensing and CI, following the
Default Workflow SOP.
What changed:
- CLAUDE.md + docs/: copied the Default Workflow so sessions load the SOP.
- state/: PROJECT, ARCHITECTURE, DECISIONS, TODO, NOTES filled in for OpenScribe.
ARCHITECTURE captures the four-part design (firmware, server, app, case) and the
three sync paths; DECISIONS records the hardware, AI-stack, storage, app and
licensing choices; TODO lays out milestones M1-M9.
- hardware/BOM.md: two build options (compact XIAO ESP32-S3 Sense; dev ESP32-S3 +
I2S mic + SD), wiring/pinout, indicative cost.
- api/openapi.yaml: the completely open API (device + server surfaces), including
recording list/download/delete and exports (wav/ogg/txt/srt/vtt/md/json).
- firmware/: PlatformIO ESP32-S3 project, two board profiles, pin map, boot scaffold
with module seams for M1-M4.
- server/: FastAPI skeleton mirroring the OpenAPI, config for self-hosted MinIO,
faster-whisper and Ollama; stub routes browsable at /docs.
- app/, case/: Flutter app plan; parametric OpenSCAD enclosure.
- Licensing: GPL-3.0 (code), CERN-OHL-S-2.0 (hardware), CC-BY-SA-4.0 (case/docs),
REUSE-style LICENSES/ with SPDX headers; LICENSING.md explains the split.
- CI: Forgejo Actions workflow builds firmware (both profiles) and lints/imports server.
Why:
- Everything self-hosted and openly licensed per the user's requirements: an open
API, three sync paths (BLE control, WiFi transfer, independent WiFi upload on
charge to generic cloud storage), and a full self-hosted transcription+summary stack.
Notes:
- No custom PCB in v1; off-the-shelf modules. Physical verification waits on parts.
- Component code is stubs at M0; features land milestone by milestone, each as its
own branch/PR per the workflow.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
