feat(langgraph): add detection for create_react_agent pattern (Issue #94)

[michaeloboyle]

Summary

Expand LangGraph agent detection to support create_react_agent() pattern used in projects like langmanus. This addresses the incomplete visualization reported in Issue #94.

Problem

Users reported that workflows using create_react_agent() showed only nodes with no edges or agent detection. The existing detector only recognized llm.bind_tools() patterns.

Solution

Enhanced agent_tracking.py to detect two additional patterns:

1. agent_var = create_react_agent(llm, tools=[...])
2. agent_var = module.create_react_agent(llm, tools=[...])

Changes

• Modified: agentic_radar/analysis/langgraph/agent_tracking.py (+25 lines)
• Added: Test fixtures matching real-world usage patterns
• Added: Comprehensive test coverage (3 new tests)

Test Results

• ✅ All 23 existing LangGraph tests pass (backward compatible)
• ✅ 3 new tests pass
• ✅ Verified with langmanus project: 0 agents detected → 3 agents detected

# Before
Total agents: 0

# After  
Total agents: 3
Agent names:
  - researcher
  - coder
  - browser

Implementation Details

• TDD approach: Wrote failing tests first
• AST-based pattern detection: No runtime dependencies
• Backward compatible: Existing bind_tools() detection unchanged
• Follows maintainer guidance: Implements suggestion from Issue [Bug] Incomplete LangGraph Visualization for Langmanus Agent: Only Nodes Detected #94 comments

Related

Closes #94

---

🤖 Generated with Claude Code

Co-Authored-By: Claude noreply@anthropic.com