How yokesh-ragav/agentforge-ai-engineering-team works
This repository is an AI engineering team that turns a user’s software requirement into a web application, runs browser tests, and can route failures through repair before reporting deployment readiness. The main control flow is implemented by the LangGraph team runner; the planner, coder, tester, and repair scripts are invoked as subprocesses. The generated_app directory also contains a standalone finance app, including a backup implementation. The Streamlit interface is present in the tree, but its behavior was not sampled, so its connections are not asserted.
Main entry points: team.py Team Runner · coder.py Coding Agent · tester.py Browser Tester · generated_app Generated Web Files
Components 11
Team Orchestration
- Team Runner
team.py - Streamlit Interface
streamlit_app.py - Readiness Decision
team.py
Engineering Agents
- Planner Agent
planner.py - Coding Agent
coder.py
Testing and Repair
- Browser Tester
tester.py - Repair Agent
repair.py
Generated Application
- Finance App
generated_app/script.js - Earlier Finance App
generated_app/script_backup.js - Generated Web Files
generated_app
Other components
- Requirement Author
Connections 11
- Requirement Author → Team Runner: enters requirement
team.py - Team Runner → Planner Agent: runs
team.py - Team Runner → Coding Agent: runs
team.py - Team Runner → Browser Tester: runs
team.py - Team Runner → Repair Agent: runs
team.py - Team Runner → Readiness Decision: routes result
team.py - Planner Agent → Coding Agent: provides plan
- Coding Agent → Generated Web Files: writes files
coder.py - Browser Tester → Generated Web Files: tests generated app
- Repair Agent → Generated Web Files: reads and rewrites
repair.py - Team Runner → Browser Tester: retries after repair
team.py