How jaisurya2823/ai-codebase-onboarding-agent works
This CLI analyzes a supplied codebase and writes a Markdown onboarding report. Its workflow scans the repository, asks an LLM to summarize structure and risks, then synthesizes an overview and first-week tasks. The graph and shared state live in the same modules; scanner helpers are deterministic filesystem logic. The README documents Groq as the inference provider. Source excerpts show the main orchestration and scanner behavior, but not every graph-node implementation in full, so the graph preserves only relationships explicitly shown or documented.
Main entry points: main.py CLI runner · graph.py LangGraph workflow · scanner.py Filesystem scanner · state.py Shared agent state
Components 11
CLI
- CLI runner
main.py
Repository scanning
- Filesystem scanner
agent/scanner.py
Analysis and reporting
- LangGraph workflow
agent/graph.py - Shared agent state
agent/state.py - Structure analysis
agent/graph.py - Deep read and risks
agent/graph.py - Report synthesis
agent/graph.py
Other components
- Developer
- Target repository
- Groq inference
- Markdown report
Connections 12
- Developer → CLI runner: runs
README.md - CLI runner → Target repository: selects for analysis
main.py - CLI runner → LangGraph workflow: builds and streams
main.py - LangGraph workflow → Filesystem scanner: scans and reads
agent/graph.py - LangGraph workflow → Structure analysis: dispatches
agent/graph.py - Structure analysis → Deep read and risks: then dispatches
agent/graph.py - Deep read and risks → Report synthesis: then dispatches
agent/graph.py - Structure analysis → Groq inference: requests analysis
agent/graph.py - Deep read and risks → Groq inference: requests risk review
agent/graph.py - Report synthesis → Groq inference: requests synthesis
agent/graph.py - LangGraph workflow → Shared agent state: shares workflow data
agent/graph.py - CLI runner → Markdown report: writes report
main.py