[Feature] Phase 4: Integrate Native AST Mapping into LangGraph Agents

[purvanshjoshi]

[Feature] Phase 4: Integrate Native AST Mapping into LangGraph Agents

Objective

With the native AST CodebaseMapper implemented (Phase 1), we must now bind it to the LangGraph execution pipeline so the agents can utilize it.

Requirements

1. Tool Registration: Inside backend/agents.py, register a new tool function: generate_architecture_map(repo_path: str). This function will instantiate the CodebaseMapper and return the AST structural summary as a JSON string or markdown block.
2. Prompt Modifications:
   • Update the Architect Agent's system prompt to instruct it to run generate_architecture_map as its primary step for codebase discovery, replacing reliance on GitNexus index sweeping.
   • Instruct the Architect to provide exact line number citations to the Implementer agent when delegating tasks (e.g., "The target class is in auth.py from line 10-45").
3. Agent State Alignment: Ensure that the generated AST summary is small enough to fit within the Groq API token context limits, or implement a chunking/summarization step before returning it to the LLM.

Acceptance Criteria

• The LangGraph state machine successfully executes generate_architecture_map.
• The Architect agent uses the new citations format to delegate implementation tasks.