mirror of https://github.com/Memo-2023/mana-monorepo.git synced 2026-05-17 21:49:40 +02:00

Till-JS 384244fe50 ♻️ refactor: migrate bots and chat-backend to mana-llm

Migrate all LLM consumers from direct Ollama calls to centralized
mana-llm service with OpenAI-compatible API.

Migrated services:
- matrix-ollama-bot
- telegram-ollama-bot
- chat-backend
- telegram-project-doc-bot

New env vars: MANA_LLM_URL, LLM_MODEL, LLM_TIMEOUT
Replaces: OLLAMA_URL, OLLAMA_MODEL, OLLAMA_TIMEOUT

2026-01-29 22:56:26 +01:00

3.4 KiB

Raw Blame History

Matrix Ollama Bot - Claude Code Guidelines

Overview

Matrix Ollama Bot provides a GDPR-compliant chat interface to local LLM inference via Ollama. It uses the Matrix protocol for messaging, which allows self-hosting all data on the Mac Mini server.

Tech Stack

Framework: NestJS 10
Matrix: matrix-bot-sdk
LLM: mana-llm service (supports Ollama + cloud providers)

Commands

# Development
pnpm install
pnpm start:dev        # Start with hot reload

# Build
pnpm build            # Production build

# Type check
pnpm type-check       # Check TypeScript types

Project Structure

services/matrix-ollama-bot/
├── src/
│   ├── main.ts               # Application entry point
│   ├── app.module.ts         # Root module
│   ├── health.controller.ts  # Health check endpoint
│   ├── config/
│   │   └── configuration.ts  # Configuration & system prompts
│   ├── bot/
│   │   ├── bot.module.ts
│   │   └── matrix.service.ts # Matrix client & command handlers
│   └── ollama/
│       ├── ollama.module.ts
│       └── ollama.service.ts # Ollama API client
├── Dockerfile
└── package.json

Matrix Commands

Command	Description
`!help`	Show help message
`!models`	List available Ollama models
`!model [name]`	Switch to a different model
`!mode [mode]`	Change system prompt mode
`!clear`	Clear chat history
`!status`	Show Ollama connection status

System Prompt Modes

Mode	Description
`default`	General assistant
`classify`	Text classification
`summarize`	Text summarization
`translate`	Translation
`code`	Programming help

Environment Variables

# Server
PORT=3311

# Matrix
MATRIX_HOMESERVER_URL=http://localhost:8008
MATRIX_ACCESS_TOKEN=syt_xxx
MATRIX_ALLOWED_ROOMS=#ollama-bot:mana.how
MATRIX_STORAGE_PATH=./data/bot-storage.json

# LLM (via mana-llm service)
MANA_LLM_URL=http://localhost:3025
LLM_MODEL=ollama/gemma3:4b
LLM_TIMEOUT=120000

Docker

# Build locally
docker build -f services/matrix-ollama-bot/Dockerfile -t matrix-ollama-bot services/matrix-ollama-bot

# Run
docker run -p 3311:3311 \
  -e MATRIX_HOMESERVER_URL=http://synapse:8008 \
  -e MATRIX_ACCESS_TOKEN=syt_xxx \
  -e MANA_LLM_URL=http://mana-llm:3025 \
  -v matrix-ollama-bot-data:/app/data \
  matrix-ollama-bot

Health Check

curl http://localhost:3311/health

Getting a Matrix Access Token

# Login to get access token
curl -X POST "https://matrix.mana.how/_matrix/client/v3/login" \
  -H "Content-Type: application/json" \
  -d '{
    "type": "m.login.password",
    "user": "ollama-bot",
    "password": "your-password"
  }'

# Response contains: {"access_token": "syt_xxx", ...}

Key Differences from Telegram Bot

Feature	Telegram	Matrix
Commands	`/command`	`!command`
Message limit	4096 chars	~65535 chars
Data storage	Telegram servers	Self-hosted
E2E encryption	Bot chats unencrypted	Optional (not enabled)
Typing indicator	`sendChatAction`	`sendTyping`

All message data stored locally on Mac Mini
No third-party data processing
Full control over data retention
Can delete all user data on request

3.4 KiB Raw Blame History