| name | astakos-ai-agent |
| description | A modular, local-first, LLM-agnostic multi-agent companion framework |
What I can accomplish
- Routine Learning: Learn user habits from conversational context and trigger proactive reminders via background scheduler.
- Hybrid Memory: Store and extract temporal (SQLite) and semantic (ChromaDB) facts, handling dedup and sentiment overrides.
- Multi-Agent Orchestration: Delegate tasks across LangGraph-based agents (Chat, Home, Web, Tech, Git, Mail, Dev).
- Smart Integrations: Control IoT devices (Vacuum), manage Google Calendar, generate/upload files to Google Drive, and parse receipts/product labels.
Required inputs
- API Keys: GEMINI_API_KEY, TELEGRAM_TOKEN
- Local Configuration:
.env setup with standard tokens
Constraints
- Local-first execution: All
.db and memory states reside strictly on the local machine.
- Environment: Python 3.11+ required.
- Approval Gates: All CRITICAL operations (e.g.
write_code, push_git) require explicit Telegram approval before execution.
Key documentation
- README.md: Full overview of the architecture and roadmap
- AGENTS.md: Coding rules and AI workflow instructions
- llms.txt: Discovery map for AI agents
Example Usage
Here is an example of how an AI Agent can use the native Astakos tools to interact with the project:
from tools.system import execute_shell_command
result = execute_shell_command("echo 'Hello Astakos'")
print(result)
