Botbot

Botbot, your not so friendly Bot

Botbot aimsto be a multi‑service chat assistant for Matrix that captures room conversations, stores them, and uses Large‑Language‑Models to provide concise summaries, answer follow‑up questions, and surface action points.

Roadmap / Aspirational goals

Persistent conversation store (PostgreSQL) for long‑range context.
Pluggable LLM back‑ends (local models, Ollama, etc.).
Structured meeting summaries with action‑items and deadlines.
Additional chat front‑ends (Telegram, WhatsApp, …).

Architecture

flowchart LR
  MS["matrix_service<br/>(Python & nio)"]
  AI["ai_service<br/>(FastAPI)"]
  Redis["Redis<br/>Deduplicates replies / guarantees idempotency"]

  MS -- "HTTP (Bearer token)" --> AI
  AI -- "Matrix Events" --> MS

  MS -.-> Redis
  AI -.-> Redis

Component	Image / Entry‑Point	Purpose
matrix_service	`python matrix_service/main.py`	Listens to Matrix rooms, forwards each message to `ai_service`, posts the reply back.
ai_service	`python ai_service/main.py` (FastAPI)	Builds a prompt, calls the configured LLM (OpenAI today), caches the reply in Redis.
redis	`redis:7`	Reply cache & simple key/value store.

The services talk to each other over the internal Docker network. Authentication between them is a static bearer token (AI_HANDLER_TOKEN).

Current Features

Auto‑Join on Invite – secure E2EE join including device verification.
Stateless AI handler – FastAPI endpoint /api/v1/message receiving a JSON payload.
Idempotent replies – duplicate Matrix events reuse the cached answer.

Quick Start (development)

# clone & cd
$ git clone https://gitea.alluna.pt/jfig/botbot.git
$ cd botbot

# copy environment template
$ cp .env.example .env
# edit .env with your homeserver, credentials and OpenAI key

# start everything (hot‑reload volumes mounted)
$ docker compose up --build

The default compose file launches three containers:

matrix_service – watches your rooms
ai_service – handles AI prompts
redis – reply cache

Stop with Ctrl‑C or docker compose down.

Production (single image per service)

Build once, then deploy with your orchestrator of choice:

$ docker compose -f docker-compose.yml --profile prod build
$ docker compose -f docker-compose.yml --profile prod up -d

Configuration

All settings are environment variables. The table below reflects the current codebase (commit ae27a2c).

Variable	Service	Default	Description
`LOG_LEVEL`	both	`INFO`	Python logging level.
`MATRIX_HOMESERVER_URL`	matrix_service	–	Matrix homeserver base URL.
`MATRIX_USER_ID`	matrix_service	–	Full user id of the bot.
`MATRIX_PASSWORD`	matrix_service	–	Password for the bot account.
`MATRIX_LOGIN_TRIES`	matrix_service	`5`	Number of login attempts before exit.
`MATRIX_LOGIN_DELAY_INCREMENT`	matrix_service	`5`	Seconds added per retry.
`AI_HANDLER_URL`	matrix_service	`http://ai_service:8000`	Where to POST messages.
`AI_HANDLER_TOKEN`	both	–	Shared bearer token (keep secret).
`OPENAI_API_KEY`	ai_service	–	Key for `openai` Python SDK.
`REDIS_URL`	ai_service	`redis://redis:6379`	Connection string used by `redis-py`.

API (ai_service)

POST /api/v1/message
Authorization: Bearer <AI_HANDLER_TOKEN>
Content-Type: application/json

{
  "roomId": "!foo:matrix.org",
  "userId": "@alice:matrix.org",
  "eventId": "$abc123",
  "serverTimestamp": 1714821630123,
  "content": "Hello there"
}

Returns {"reply": "Hello Alice!"} or HTTP 401/500 on error.

Contributing

Pull requests are welcome! Please open an issue first to discuss what you want to change. All source code is formatted with ruff / black – run pre-commit run --all-files before pushing.

License

MIT

5.0 KiB Raw Permalink Blame History Unescape Escape