+ AI Agents
Multi-Crew AI Agent System
Five specialised CrewAI crews (debate, financial research, stock picking, code generation, software engineering) with web search, memory and sandboxed code execution.
Source code is private
Overview
A CrewAI project with five independent crews that each solve a different multi-step task. Crews use sequential and manager-led processes, web search via Serper, long-term memory in SQLite and push notifications. The code-generation crew runs generated code in isolated Docker containers with timeouts and retries, and outputs are produced as structured Markdown, JSON or code files.
Highlights
- 5 specialised crews: debate, financial research, stock picking, code generation and software engineering
- Sequential and manager-based (hierarchical) agent processes
- Safe code execution in isolated Docker containers with timeout and retry
- Long-term memory in SQLite to avoid repeated picks
- Structured outputs (Markdown reports, JSON, code) and push notifications