🎤 Daily Meeting Transcriber

A FastAPI service to transcribe and summarize your daily meetings using OpenAI Whisper and GPT-4o mini, with optional Notion integration.

---

Table of Contents

• Features
• Quickstart
• Notion Integration
• Custom Prompts
• OpenAI-compatible platforms
• Project Structure
• Deployment
• Contributing
• License

---

Features

• 🎧 Audio transcription via OpenAI Whisper (mp3, m4a, wav, and any format supported by ffmpeg)
• 🤖 Structured Markdown summary generation via GPT-4o mini
• 📦 Optional Notion integration: automatically publish summaries to your Notion database
• 🌍 Multilingual — language is configurable via PROMPT_CONFIG
• 🪄 FastAPI HTTP API with interactive OpenAPI docs at /docs
• 📊 Token usage tracking in every response
• 🦄 Structured logging with Loguru

---

Quickstart

1. Clone & Install

git clone https://github.com/mael-app/daily-meeting-transcriber.git
cd daily-meeting-transcriber
python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate
pip install -r requirements.txt

ffmpeg must be installed on your system for audio processing (brew install ffmpeg / apt install ffmpeg).

2. Configure environment variables

Copy .env.example to .env and fill in your values:

cp .env.example .env

Variable	Required	Description
OPENAI_API_KEY	Yes	API key for your OpenAI-compatible platform
PROMPT_CONFIG	Yes	Prompts as a JSON string (see Custom Prompts)
NOTION_TOKEN	No	Notion integration token
NOTION_DB_SCHEMA	No*	Notion database schema as a JSON string
NOTION_TITLE	No	Title for created Notion pages (default: Daily)
NOTION_CATEGORY	No	Category select value for Notion pages (default: Standup)
OPENAI_WHISPER_ENDPOINT	No	Whisper-compatible transcription endpoint (default: OpenAI)
OPENAI_CHAT_ENDPOINT	No	Chat completions endpoint (default: OpenAI)
WHISPER_MODEL	No	Transcription model (default: whisper-1)
CHAT_MODEL	No	Chat model (default: gpt-4o-mini)

*Required only if NOTION_TOKEN is set and no schema is uploaded per-request.

3. Run locally

# With hot reload
python -m uvicorn app:app --reload --port 8080

# Or with Docker Compose (loads .env automatically)
docker-compose up --build

4. API Usage

Open http://localhost:8080/docs for interactive documentation.

Endpoints:

• GET /health — Healthcheck
• POST /process-audio — Transcribe and summarize an audio file

Example:

curl -X POST "http://localhost:8080/process-audio" \
  -F "file=@meeting.m4a"

Response:

{
  "tokens": 1234,
  "transcript_success": true,
  "markdown": "### Work from yesterday\n- ...",
  "notion_sent": false
}

---

Notion Integration

To enable Notion publishing, set NOTION_TOKEN and NOTION_DB_SCHEMA.

If NOTION_TOKEN is not set, Notion is skipped entirely and the response still contains the transcript and summary.

1. Create a Notion integration

Go to https://www.notion.so/profile/integrations, create a new integration, and copy the Internal Integration Token — this is your NOTION_TOKEN.

2. Get your database ID

Open your Notion database in the browser. The URL looks like:

https://www.notion.so/yourworkspace/xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx?v=...

The 32-character string before the ? is your database ID.

Alternatively, use the Notion API to retrieve it:

curl -X GET "https://api.notion.com/v1/search" \
  -H "Authorization: Bearer secret_..." \
  -H "Notion-Version: 2022-06-28" \
  -H "Content-Type: application/json" \
  -d '{"filter": {"value": "database", "property": "object"}}'

Look for the "id" field in the response matching your database.

3. Connect the integration to your database

In Notion, open the database → click ... (top right) → Connections → add your integration.

4. Set the environment variables

export NOTION_TOKEN="secret_..."
export NOTION_DB_SCHEMA='{"id": "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx"}'

The NOTION_DB_SCHEMA value only needs to contain the id (or database_id) key. The service reads that field to identify the target database.

You can also upload the schema per-request via the optional notion_schema field:

curl -X POST "http://localhost:8080/process-audio" \
  -F "file=@meeting.m4a" \
  -F "notion_schema=@schema.json"

---

Custom Prompts

PROMPT_CONFIG is required. Set it to a JSON string containing your prompts:

export PROMPT_CONFIG='{
  "system_prompt": "You are an assistant that summarizes developer standups.",
  "user_prompt": "Summarize the following transcript:\n---\n{transcript}\n---",
  "language": "en"
}'

The {transcript} placeholder is required in user_prompt. The language value is passed directly to the Whisper API.

On Kubernetes, use a ConfigMap:

apiVersion: v1
kind: ConfigMap
metadata:
  name: transcriber-config
data:
  PROMPT_CONFIG: |
    {"system_prompt": "...", "user_prompt": "...\n---\n{transcript}\n---", "language": "en"}

---

OpenAI-compatible platforms

OPENAI_API_KEY, OPENAI_WHISPER_ENDPOINT, and OPENAI_CHAT_ENDPOINT work with any platform that exposes an OpenAI-compatible API — Azure OpenAI, Together AI, local inference servers, etc.

Example with a local server:

OPENAI_API_KEY="local"
OPENAI_WHISPER_ENDPOINT="http://localhost:8000/v1/audio/transcriptions"
OPENAI_CHAT_ENDPOINT="http://localhost:8000/v1/chat/completions"
WHISPER_MODEL="whisper-large-v3"
CHAT_MODEL="llama-3.1-8b-instruct"

---

Project Structure

app.py              FastAPI entrypoint
services/
  audio_service.py      Request orchestrator
  transcription_service.py  Whisper API integration
  summary_service.py    GPT-4o mini integration
  notion_service.py     Notion API integration
utils/
  config.py         App configuration loader
  env.py            Environment variable helper

---

Deployment

The CI/CD pipeline (.github/workflows/docker-publish.yml) builds a multi-arch Docker image (linux/amd64, linux/arm64) and pushes it to GitHub Container Registry (ghcr.io) on every push to main.

---

Contributing

See CONTRIBUTING.md.

---

License

See LICENSE.