Kimi Project Intelligence MCP (Kimi-PIMCP) v0.2.0 🚀

A Model Context Protocol (MCP) server for Kimi-CLI that provides semantic indexing, intelligent context retrieval, and dynamic skill selection for enhanced code assistance.

✨ Features

Core Features

• Semantic Indexing - Index projects using ML embeddings (sentence-transformers)
• Intelligent Retrieval - Find relevant code with cosine similarity + MMR for diversity
• Skill Routing - Automatically detect user intent and select optimal skills
• Caveman Compression - Heuristic-based text compression for token optimization
• MCP Protocol - Full compatibility with Kimi-CLI via stdio JSON-RPC

New in v0.2.0

• Incremental Indexing - Only reindex changed files (10x faster reindexing)
• Query Caching - LRU cache for frequently used queries
• File Watcher - Auto-reindex when files change
• Dependency Graph - Analyze import relationships between files
• Git Integration - Index only modified files vs HEAD
• Code Summarization - Auto-summarize code chunks
• Similar Code Detection - Find duplicate/similar code blocks
• REST API - HTTP API for external integrations
• Web UI - Beautiful web interface for managing indexes

📁 Architecture

kimi-pimcp/
├── src/
│   ├── indexer.py           # Semantic indexing with embeddings (OPTIMIZED)
│   ├── retriever.py         # Context retrieval with MMR + caching
│   ├── compressor.py        # Caveman text compression with tiktoken
│   ├── server.py            # MCP stdio server (multi-project support)
│   ├── file_watcher.py      # File change monitoring
│   ├── dependency_graph.py  # Import/dependency analysis
│   ├── git_integration.py   # Git integration for incremental updates
│   ├── code_summarizer.py   # Code summarization and duplicate detection
│   ├── rest_api.py          # HTTP REST API
│   ├── web_ui.py            # Web interface
│   └── skills/
│       ├── base.py          # Base skill classes
│       ├── router.py        # Intent classification
│       └── prompts/         # System prompts
├── data/datasets/           # Training datasets
├── tests/                   # Unit tests
├── notebooks/               # Validation notebooks
└── docs/                    # Documentation

🚀 Installation

Linux/Mac

chmod +x install.sh
./install.sh

Windows

.\install.ps1

Manual Installation

# Clone the repository
git clone <repository-url>
cd kimi-pimcp

# Create virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Pre-download models
python -c "from sentence_transformers import SentenceTransformer; SentenceTransformer('sentence-transformers/all-MiniLM-L6-v2')"

💻 Usage

Running the MCP Server

# Start the MCP server
python src/server.py

# Or use the installed command
kimi-pimcp

The server communicates via stdin/stdout using JSON-RPC 2.0 protocol.

Running the REST API

# Start the REST API server
python src/rest_api.py --host 0.0.0.0 --port 8000

# With auto-reload (development)
python src/rest_api.py --reload

The API will be available at http://localhost:8000

Running the Web UI

# Start the web UI server
python src/web_ui.py

Then open http://localhost:8080 in your browser.

🛠️ Available Tools

Core Tools

1. Initialize Index

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "initialize_index",
    "arguments": {
      "project_path": "/path/to/project",
      "force_reindex": false
    }
  }
}

2. Query Context

{
  "jsonrpc": "2.0",
  "id": 2,
  "method": "tools/call",
  "params": {
    "name": "query_context",
    "arguments": {
      "query": "authenticate user",
      "top_k": 5,
      "filter_ext": [".py", ".js"],
      "use_mmr": true
    }
  }
}

3. Select Skill

{
  "jsonrpc": "2.0",
  "id": 3,
  "method": "tools/call",
  "params": {
    "name": "select_skill",
    "arguments": {
      "query": "fix login bug"
    }
  }
}

4. Compress Output

{
  "jsonrpc": "2.0",
  "id": 4,
  "method": "tools/call",
  "params": {
    "name": "compress_output",
    "arguments": {
      "text": "Please help me fix this bug...",
      "level": "full"
    }
  }
}

New Tools in v0.2.0

5. Get Query History

{
  "jsonrpc": "2.0",
  "id": 5,
  "method": "tools/call",
  "params": {
    "name": "get_query_history",
    "arguments": {
      "limit": 10
    }
  }
}

6. Switch Project

{
  "jsonrpc": "2.0",
  "id": 6,
  "method": "tools/call",
  "params": {
    "name": "switch_project",
    "arguments": {
      "project_path": "/path/to/other/project"
    }
  }
}

7. Get Dependencies

{
  "jsonrpc": "2.0",
  "id": 7,
  "method": "tools/call",
  "params": {
    "name": "get_dependencies",
    "arguments": {
      "filepath": "/path/to/file.py"
    }
  }
}

8. Find Similar Code

{
  "jsonrpc": "2.0",
  "id": 8,
  "method": "tools/call",
  "params": {
    "name": "find_similar_code",
    "arguments": {
      "filepath": "/path/to/file.py"
    }
  }
}

9. Summarize Chunk

{
  "jsonrpc": "2.0",
  "id": 9,
  "method": "tools/call",
  "params": {
    "name": "summarize_chunk",
    "arguments": {
      "content": "def hello(): print('world')",
      "chunk_type": "function",
      "language": "python"
    }
  }
}

10. Export/Import Index

{
  "jsonrpc": "2.0",
  "id": 10,
  "method": "tools/call",
  "params": {
    "name": "export_index",
    "arguments": {
      "output_path": "/path/to/export"
    }
  }
}

🎯 Skills

Skill	Description	Use Cases
debugger	Debug code issues	Errors, exceptions, crashes
architect	System design	Architecture, patterns, scalability
explainer	Explain code/concepts	Documentation, understanding
tester	Testing & QA	Unit tests, integration tests
caveman	Concise responses	Token-optimized output

⚙️ Configuration

Edit config.yaml to customize:

models:
  embedding_model: "sentence-transformers/all-MiniLM-L6-v2"
  use_cross_encoder: true

indexer:
  supported_extensions: [".py", ".js", ".ts"]
  exclude_dirs: ["node_modules", ".git"]
  max_file_size_mb: 10

compressor:
  default_level: "auto"

performance:
  query_cache_size: 100
  model_cache_size: 2

📊 Performance Targets

Operation	Target	v0.1.0	v0.2.0
Indexing	<100ms/file	~50ms	~30ms (incremental)
Query	<200ms	~100ms	~50ms (cached)
Classification	<10ms	~5ms	~3ms
Compression	<5ms	~2ms	~1ms
Memory	<500MB	~300MB	~250MB

🧪 Testing

# Run unit tests
python -m pytest tests/ -v

# Run with coverage
python -m pytest tests/ --cov=src --cov-report=html

# Run validation notebook
jupyter notebook notebooks/validation.ipynb

📚 REST API Endpoints

Endpoint	Method	Description
/health	GET	Health check
/index	POST	Index a project
/index/status	GET	Get indexing status
/query	POST	Search code
/compress	POST	Compress text
/skills/select	POST	Detect skill
/skills	GET	List skills
/stats	GET	Get all stats
/dependencies	GET	Get file dependencies
/dependencies/stats	GET	Dependency graph stats
/git/status	GET	Git status
/git/changes	GET	Git changes
/cache/clear	POST	Clear all caches

📖 API Reference

Indexer Module

from indexer import ProjectIndexer, get_indexer

indexer = get_indexer(project_path="/path/to/project")
stats = indexer.index_project("/path/to/project")
print(f"Indexed {stats['files_indexed']} files")
print(f"Skipped {stats['files_skipped']} unchanged files")

Retriever Module

from retriever import ContextRetriever, get_retriever

retriever = get_retriever()
retriever.load_index("/path/to/project")
results = retriever.query("authenticate user", top_k=5)

# Check cache stats
print(retriever.get_stats()['cache'])

File Watcher

from file_watcher import ProjectFileWatcher

def on_change(event):
    print(f"File {event.change_type.value}: {event.filepath}")

watcher = ProjectFileWatcher("/path/to/project", on_change)
watcher.start()

Dependency Graph

from dependency_graph import DependencyGraph

graph = DependencyGraph("/path/to/project")
graph.build()

# Get dependencies
deps = graph.get_dependencies("/path/to/file.py")
dependents = graph.get_dependents("/path/to/file.py")

# Find circular dependencies
cycles = graph.find_circular_dependencies()

Git Integration

from git_integration import GitIntegration

git = GitIntegration("/path/to/project")

# Get changed files since last commit
changes = git.get_changed_files_since("HEAD~1")
files_to_index, files_to_remove = git.get_files_to_index("HEAD~1")

🤝 Contributing

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Make your changes
4. Add tests
5. Submit a pull request

📄 License

MIT License - see LICENSE file for details.

🙏 Acknowledgments

• sentence-transformers for embeddings
• FAISS for vector search
• scikit-learn for classification
• watchdog for file watching
• FastAPI for REST API