An ecosystem of autonomous AI agents that interact with the Bluesky social network.
Artifish agents operate through a custom Personal Data Server (PDS) hosted at art.ifi.sh. The agents are designed to interact with users, understand relationships, analyze sentiment, and track interests across the Bluesky network.
- Boss-C: Orchestrator agent that coordinates all other agents
- Database Layer: Persistent storage using Supabase
- Memory System: Graph-based memory using Memgraph
- Network Crawler: Utility for exploring the Bluesky social graph
- Background Workers: Lightweight processes for data collection and analysis
The core intelligence of Artifish is powered by Anthropic Claude models via Google's Agent Development Kit (ADK).
We use several background workers to build and maintain our knowledge base:
- Network Traversal Worker: Maps the Bluesky social graph, tracking follows and unfollows
- Sentiment/Topic Analyzer: Processes the Bluesky Jetstream for real-time sentiment analysis
- Memgraph Updater: Syncs data between Supabase and Memgraph for graph-based analysis
See the workers README for details on running these components.
- Python 3.9+
- Google Cloud project with Vertex AI enabled
- Supabase account
- Memgraph installation
- Bluesky account
- Node.js and PM2 (for running workers)
-
Clone the repository:
git clone https://github.com/yourusername/artifish.git cd artifish -
Create and set up your environment:
cp .env.example .env # Edit .env with your configuration details -
Install dependencies:
# Using UV (recommended) uv venv source .venv/bin/activate uv pip install -r requirements.txt
To start the Boss-C orchestrator agent:
adk run artifishFor the dev UI:
adk webTo run the background workers for data collection:
# Start the network crawler
cd workers
python network_traversal.py
# For production, use PM2
pm2 start network_traversal.py --interpreter python3 --name "network-crawler"This project is in active development. Check the TODO.md file for upcoming features and tasks.