-
Notifications
You must be signed in to change notification settings - Fork 0
Home
🧠 Adaptive Neural Knowledge System for AI Code Understanding
Achieve 40-70x token reduction when working with AI coding assistants
NeuralMind is a Python library that creates intelligent, query-aware context from your codebase knowledge graphs. When working with AI coding assistants like Claude, GPT-4, or Cursor, context window limits are a fundamental challenge. Loading an entire codebase consumes 50,000+ tokens, leaving little room for meaningful conversation.
NeuralMind solves this by implementing a 4-layer progressive disclosure architecture that loads only what's relevant to your query—achieving massive token savings while maintaining deep understanding of your code.
| Page | Description |
|---|---|
| Installation | Detailed setup instructions for all platforms |
| CLI Reference | Complete command-line interface documentation |
| API Reference | Python API documentation with examples |
| Architecture | Deep dive into the 4-layer system |
| Integration Guide | MCP, graphify, and tool integrations |
| Troubleshooting | Common issues and solutions |
40-70x fewer tokens than loading full codebases. Turn 50,000+ token contexts into ~1,000 tokens.
Intelligent layering system that loads context progressively:
- L0: Identity (~100 tokens) - Project name, description
- L1: Summary (~500 tokens) - Architecture overview
- L2: On-Demand (~200-500 tokens) - Query-relevant modules
- L3: Search (~200-500 tokens) - Semantic search results
Find code entities by meaning, not just keywords. Powered by ChromaDB embeddings.
Understands logical code clusters and their relationships through graph-based analysis.
Model Context Protocol server for seamless integration with Claude Desktop, Cursor, and other MCP-compatible tools.
Only re-embed changed nodes for fast rebuilds when your code changes.
# Install NeuralMind
pip install neuralmind
# Generate knowledge graph (requires graphify)
pip install graphifyy
graphify update /path/to/project
# Build neural index
neuralmind build /path/to/project
# Query your codebase
neuralmind query /path/to/project "How does authentication work?"| Project | Nodes | Communities | Avg Tokens | Reduction |
|---|---|---|---|---|
| cmmc20 | 241 | 93 | 765 | 65.6x |
| mempalace | 1,626 | 34 | 1,089 | 46.0x |
- GitHub Issues: Report bugs or request features
- Discussions: Ask questions and share ideas
- Contributing: See our Contributing Guide
NeuralMind is released under the MIT License. See LICENSE for details.