Miana — Discord Meeting Recorder Bot

A Discord bot that joins voice channels, records meetings, transcribes Urdu/English speech via OpenAI Whisper, and generates AI-powered meeting notes using GPT-4o-mini.

Features

• /record — Join a voice channel and start recording (per-user audio streams)
• /stop — Stop recording and process the audio
• /status — Check current recording status
• /transcript <session_id> — Retrieve a previous session's info
• /delete <session_id> — Delete audio and transcript data
• Urdu/English/mixed language support
• Speaker-labeled transcript with timestamps
• AI-generated: summary, key decisions, action items, discussion points, risks/blockers
• Output formats: .md, .txt, .json
• Automatic audio cleanup after 24 hours

Architecture

bot/        TypeScript + discord.js + @discordjs/voice
api/        Python FastAPI — session management + transcript retrieval
worker/     Python — audio processing + transcription + summarization

Services communicate via:

• Redis (BullMQ) — job queue between bot and worker
• PostgreSQL — session metadata, participants, transcript segments, action items
• Shared volume — audio files (Railway volume)

Quick Start (Local)

Prerequisites

• Node.js 20+
• Python 3.11+
• Docker + Docker Compose
• FFmpeg
• A Discord bot token (Discord Developer Portal)
• An OpenAI API key

1. Clone and configure

cp .env.example .env
# Edit .env with your DISCORD_TOKEN, DISCORD_CLIENT_ID, OPENAI_API_KEY, etc.

2. Start infrastructure

docker compose up postgres redis -d

3. Register slash commands

cd apps/bot
npm install
npm run deploy-commands

4. Start services

# Terminal 1 — Bot
npm run bot

# Terminal 2 — API
cd apps/api
pip install -e .
uvicorn app.main:app --reload

# Terminal 3 — Worker
cd apps/worker
pip install -e .
python -m app.worker

Railway Deployment

1. Push this repo to GitHub
2. Create a new Railway project → Deploy from GitHub repo
3. Add services:
   • bot → Dockerfile: docker/bot.Dockerfile
   • api → Dockerfile: docker/api.Dockerfile
   • worker → Dockerfile: docker/worker.Dockerfile
4. Add Railway addons:
   • PostgreSQL
   • Redis
   • Volume (mount at /data/audio for api + worker)
5. Set environment variables for each service (see .env.example)
6. Deploy

Discord Bot Permissions

Required bot permissions (OAuth2 scopes):

• bot + applications.commands
• Bot permissions: Send Messages, Embed Links, Attach Files, Connect, Speak, Use Voice Activity

Required gateway intents:

• GUILDS, GUILD_VOICE_STATES, GUILD_MESSAGES, MESSAGE_CONTENT

Environment Variables

Variable	Description	Required
DISCORD_TOKEN	Bot token from Discord Developer Portal	✅
DISCORD_CLIENT_ID	Application client ID	✅
DISCORD_GUILD_ID	Test guild ID for instant command registration	Optional
OPENAI_API_KEY	OpenAI API key for Whisper + GPT	✅
DATABASE_URL	PostgreSQL connection URL	✅
REDIS_URL	Redis connection URL	✅
INTERNAL_API_SECRET	Shared secret between bot and API	✅
API_BASE_URL	URL of the API service	✅
AUDIO_STORAGE_PATH	Where to store audio files	Optional
AUDIO_RETENTION_HOURS	Hours to keep raw audio (default: 24)	Optional

Cost Estimate

Service	Cost
Railway (bot + api + worker)	~$5–10/month
OpenAI Whisper	~$0.006/min of audio
OpenAI GPT-4o-mini	~$0.01–0.05 per meeting
Typical 30-min meeting	~$0.20–0.25

Privacy

• The bot posts a visible consent message before recording begins
• Raw audio is deleted after 24 hours by default
• Users can delete their session at any time with /delete
• The bot never joins a channel without an explicit /record command