Lila MCP Server

#!/usr/bin/env python3 """Production Repository Analyzer using Claude Agent SDK. This script performs comprehensive repository analysis with: - Incremental file outputs for each phase - Full progress visibility (tool uses and results) - Structured output directory - Phase-based analysis with checkpoints """ import asyncio from pathlib import Path from claude_agent_sdk import ( AgentDefinition, AssistantMessage, ClaudeAgentOptions, ClaudeSDKClient, ResultMessage, TextBlock, ToolResultBlock, ToolUseBlock, UserMessage, ) # Output directory structure OUTPUT_DIR = Path("repo_analysis") DOCS_DIR = OUTPUT_DIR / "docs" DIAGRAMS_DIR = OUTPUT_DIR / "diagrams" REPORTS_DIR = OUTPUT_DIR / "reports" def display_message(msg, show_tools=True): """Display message content with full visibility into tool usage.""" if isinstance(msg, AssistantMessage): for block in msg.content: if isinstance(block, TextBlock): print(f"\n🤖 Claude: {block.text}") elif isinstance(block, ToolUseBlock) and show_tools: print(f"\n🔧 Using tool: {block.name}") if block.name in ["Read", "Grep", "Glob"]: # Show what files are being analyzed if "file_path" in block.input: print(f" Reading: {block.input['file_path']}") if "pattern" in block.input: print(f" Pattern: {block.input['pattern']}") elif block.name == "Write": # Show what files are being created if "file_path" in block.input: print(f" ✍️ Writing: {block.input['file_path']}") elif block.name == "Bash": # Show commands being executed if "command" in block.input: print(f" Command: {block.input['command']}") elif isinstance(msg, UserMessage): for block in msg.content: if isinstance(block, ToolResultBlock) and show_tools: # Show results of tool operations content = str(block.content)[:200] if block.content else "None" print(f" ✅ Result: {content}...") elif isinstance(msg, ResultMessage): print("\n" + "=" * 70) print("✅ Phase completed") if msg.total_cost_usd and msg.total_cost_usd > 0: print(f"💰 Cost: ${msg.total_cost_usd:.4f}") print("=" * 70) async def phase_1_component_inventory(client: ClaudeSDKClient): """Phase 1: Generate component inventory.""" print("\n" + "=" * 70) print("📋 PHASE 1: Component Inventory") print("=" * 70) await client.query( f"""Use the analyzer agent to create a comprehensive component inventory. Analyze the codebase and document: 1. All Python modules and their purposes 2. Key classes and functions with descriptions 3. Public API surface vs internal implementation 4. Entry points and main interfaces Write your analysis to: {DOCS_DIR}/01_component_inventory.md Use this structure: # Component Inventory ## Public API [List public modules, classes, functions] ## Internal Implementation [List internal modules and their roles] ## Entry Points [Document main entry points] Include file paths and line numbers for all references.""" ) async for msg in client.receive_response(): display_message(msg) async def phase_2_architecture_diagrams(client: ClaudeSDKClient): """Phase 2: Generate architecture diagrams.""" print("\n" + "=" * 70) print("🏗️ PHASE 2: Architecture Diagrams") print("=" * 70) await client.query( f"""Use the analyzer agent to create architecture diagrams. Generate Mermaid diagrams showing: 1. System architecture (layered view) 2. Component relationships 3. Class hierarchies 4. Module dependencies Write your diagrams to: {DIAGRAMS_DIR}/02_architecture_diagrams.md Use this structure: # Architecture Diagrams ## System Architecture ```mermaid [System architecture diagram] ``` ## Component Relationships ```mermaid [Component relationship diagram] ``` ## Class Hierarchies ```mermaid [Class hierarchy diagrams] ``` Include explanations for each diagram.""" ) async for msg in client.receive_response(): display_message(msg) async def phase_3_data_flows(client: ClaudeSDKClient): """Phase 3: Document data flows.""" print("\n" + "=" * 70) print("🔄 PHASE 3: Data Flow Analysis") print("=" * 70) await client.query( f"""Use the analyzer agent to document data flows. Create sequence diagrams showing: 1. Simple query flow 2. Interactive client session flow 3. Tool permission callback flow 4. MCP server communication flow 5. Message parsing and routing Write your analysis to: {DOCS_DIR}/03_data_flows.md Use this structure: # Data Flow Analysis ## Query Flow ```mermaid [Sequence diagram] ``` [Explanation] ## Interactive Session Flow ```mermaid [Sequence diagram] ``` [Explanation] [Continue for other flows...]""" ) async for msg in client.receive_response(): display_message(msg) async def phase_4_api_documentation(client: ClaudeSDKClient): """Phase 4: Generate API documentation.""" print("\n" + "=" * 70) print("📚 PHASE 4: API Documentation") print("=" * 70) await client.query( f"""Use the doc-writer agent to create comprehensive API documentation. Document: 1. All public functions and classes 2. Parameters, return types, and examples 3. Usage patterns and best practices 4. Configuration options Write your documentation to: {DOCS_DIR}/04_api_reference.md Use clear examples and link to source files.""" ) async for msg in client.receive_response(): display_message(msg) async def phase_5_synthesis(client: ClaudeSDKClient): """Phase 5: Create final synthesis document.""" print("\n" + "=" * 70) print("📖 PHASE 5: Final Synthesis") print("=" * 70) await client.query( f"""Use the doc-writer agent to create a comprehensive README. Review all previously generated documents in: - {DOCS_DIR}/01_component_inventory.md - {DIAGRAMS_DIR}/02_architecture_diagrams.md - {DOCS_DIR}/03_data_flows.md - {DOCS_DIR}/04_api_reference.md Create a synthesis document at: {OUTPUT_DIR}/README.md Structure: # Repository Architecture Documentation ## Overview [High-level summary] ## Quick Start [How to use this documentation] ## Architecture Summary [Key insights from diagrams] ## Component Overview [Summary of components] ## Data Flows [Key flow patterns] ## References [Links to detailed docs] Make it comprehensive but accessible.""" ) async for msg in client.receive_response(): display_message(msg) async def verify_outputs(): """Verify all expected outputs were created.""" print("\n" + "=" * 70) print("🔍 Verifying Outputs") print("=" * 70) expected_files = [ DOCS_DIR / "01_component_inventory.md", DIAGRAMS_DIR / "02_architecture_diagrams.md", DOCS_DIR / "03_data_flows.md", DOCS_DIR / "04_api_reference.md", OUTPUT_DIR / "README.md", ] for file_path in expected_files: if file_path.exists(): size = file_path.stat().st_size print(f"✅ {file_path} ({size:,} bytes)") else: print(f"❌ {file_path} - NOT FOUND") print("=" * 70) async def main(): """Run comprehensive repository analysis in phases.""" print("\n" + "=" * 70) print("🚀 Claude Agent SDK Repository Analyzer") print("=" * 70) print(f"\nOutput directory: {OUTPUT_DIR.absolute()}") print(f"- Documentation: {DOCS_DIR}") print(f"- Diagrams: {DIAGRAMS_DIR}") print(f"- Reports: {REPORTS_DIR}") print("\nThis will generate comprehensive repository documentation") print("in incremental phases with progress visibility.") print("=" * 70) # Define specialized agents analyzer_agent = AgentDefinition( description="Analyzes code structure, patterns, and architecture", prompt="""You are a code analyzer expert. Your job is to: 1. Examine code structure, patterns, and architecture systematically 2. Generate clear Mermaid diagrams for visualization 3. Write comprehensive documentation with examples 4. Reference specific files and line numbers 5. Create well-structured markdown documents IMPORTANT: When asked to write to a file, ALWAYS use the Write tool to create the actual file. Do not just describe what you would write. Be thorough but concise. Focus on clarity and accuracy.""", tools=["Read", "Grep", "Glob", "Write", "Bash"], model="sonnet", ) doc_writer_agent = AgentDefinition( description="Writes comprehensive technical documentation", prompt="""You are a technical documentation expert. Your job is to: 1. Write clear, comprehensive documentation with examples 2. Create well-organized markdown documents 3. Include diagrams where helpful 4. Focus on developer experience and clarity 5. Link to source files with specific line numbers IMPORTANT: When asked to write to a file, ALWAYS use the Write tool to create the actual file. Do not just describe what you would write. Make documentation accessible and practical.""", tools=["Read", "Write", "Grep", "Glob"], model="sonnet", ) options = ClaudeAgentOptions( agents={"analyzer": analyzer_agent, "doc-writer": doc_writer_agent}, allowed_tools=["Read", "Write", "Grep", "Glob", "Bash"], permission_mode="acceptEdits", # Auto-approve file writes cwd=".", # Current directory ) try: async with ClaudeSDKClient(options=options) as client: # Run analysis in phases await phase_1_component_inventory(client) await phase_2_architecture_diagrams(client) await phase_3_data_flows(client) await phase_4_api_documentation(client) await phase_5_synthesis(client) # Verify all outputs await verify_outputs() print( f"\n✅ Analysis complete! View results in: {OUTPUT_DIR.absolute()}/README.md" ) except Exception as e: print(f"\n❌ Error during analysis: {e}") print("\nPartial results may be available in:", OUTPUT_DIR.absolute()) raise if __name__ == "__main__": asyncio.run(main())