Skip to content

slackbot

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History

slackbot

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
src/slackbot
src/slackbot
 
 
.dockerignore
.dockerignore
 
 
CLAUDE.md
CLAUDE.md
 
 
Dockerfile.slackbot
Dockerfile.slackbot
 
 
README.md
README.md
 
 
pyproject.toml
pyproject.toml
 
 

README.md

Marvin Slackbot

A Slack chatbot powered by AI (GPT-5 or Claude) with memories and Prefect-specific knowledge.

Project Structure

├── api.py         # FastAPI app and Slack event handlers
├── core.py        # Database, agent, and memory management
├── settings.py    # Configuration management
└── __main__.py    # Entry point

Setup

The slackbot can be run locally with minimal setup.

Local Development

# Install dependencies (from the repo root)
uv sync --extra slackbot

Configuration

Create a .env file in your project directory:

# Required Prefect Secrets (configured via UI or CLI)
# - test-slack-api-token          # Bot User OAuth Token  
# - openai-api-key                # For OpenAI models (if using GPT-5)
# - anthropic-api-key             # For Claude models
# - marvin-slackbot-github-token  # For searching GitHub issues
# - tpuf-api-key                  # TurboPuffer API key for vector storage

# Required Prefect Variables (configured via UI or CLI)
# - marvin_ai_model               # Default model to use (e.g., "gpt-5", "claude-sonnet-4-6")
# - marvin_bot_model              # Optional override for the Slackbot response model
# - marvin_memory_synthesis_model # Optional override for memory/profile synthesis (default: claude-haiku-4-5-20251001)
# - admin-slack-id                # Slack user ID for admin notifications

# Optional Settings (with MARVIN_SLACKBOT_ prefix)
MARVIN_SLACKBOT_TEST_MODE=true                # Enable auto-reload for development
MARVIN_SLACKBOT_HOST=0.0.0.0                  # Server host
MARVIN_SLACKBOT_PORT=4200                     # Server port
MARVIN_SLACKBOT_LOG_LEVEL=INFO                # Logging level
MARVIN_SLACKBOT_SLACK_API_TOKEN=xoxb-...      # Slack bot token (or use test-slack-api-token secret)
MARVIN_SLACKBOT_MAX_TOOL_CALLS_PER_TURN=50    # Max tool calls per agent turn (default: 50)
MARVIN_SLACKBOT_USER_MESSAGE_MAX_TOKENS=500   # Max tokens in user messages (default: 500)
MARVIN_SLACKBOT_TEMPERATURE=0.2               # Model temperature (default: 0.2, auto-set to 1.0 for GPT-5)

# Vector Store (optional, will use tpuf-api-key secret if not set)
TURBOPUFFER_API_KEY=abcd1234       # For vectorstore queries and storing user context

Slack App Setup

  1. Create a new Slack app at https://api.slack.com/apps
  2. Add a bot user with required scopes:
    • app_mentions:read
    • channels:read
    • chat:write
    • groups:read
    • im:read
    • mpim:read
  3. Set up event subscriptions:
    • URL: https://{YOUR_DOMAIN}/chat
    • Subscribe to bot events: app_mention, team_join

Running Locally

  1. Start ngrok in one terminal:
ngrok http 4200  # Or your configured port
  1. Start the bot in another terminal:
uv run --extra slackbot -m slackbot

Testing

Mention the bot in any channel it's invited to:

@Marvin What's new in Prefect?

The bot will:

  • Search Prefect documentation
  • Look through GitHub issues
  • Remember previous interactions
  • Provide context-aware responses

Model Configuration

The bot supports both OpenAI (GPT-5) and Anthropic (Claude) models. Configure via the marvin_ai_model Prefect Variable:

  • gpt-5: Latest OpenAI model (temperature automatically set to 1.0)
  • claude-sonnet-4-6: Current default Claude model
  • Any other supported model name from either provider

Channel Restrictions

The bot can be configured to only respond in designated channels per workspace. Users mentioning the bot in other channels will receive a redirect message. The workspace-to-channel mapping is configured in _internal/constants.py:

WORKSPACE_TO_CHANNEL_ID = {
    "TL09B008Y": "C04DZJC94DC",  # Prefect Community -> #ask-marvin
    "TAN3D79AL": "C046WGGKF4P",  # Prefect -> #ask-marvin-tests
    # Add more workspace mappings as needed
}

Development Features

  • Auto-reload in test mode
  • Colored logging output
  • SQLite message history
  • TurboPuffer vector storage for user context
  • Configurable via environment variables or .env file

Production Deployment

For deploying to Cloud Run or similar services, refer to: