Context Engine

About

Context Engine is a comprehensive, plug-and-play retrieval stack designed to accelerate the development and deployment of context-aware AI agents. It integrates code indexing, hybrid search, and optional local LLM decoding (via llama.cpp or GLM) into a unified solution. By automating micro-chunking, token budgeting, and live reindexing, it ensures AI agents always have access to precise, up-to-date context from large or fast-changing codebases, resolving issues like fragmented infrastructure and context drift. The platform provides a consistent stack for memory, search, and decoding, enabling product and DevEx teams to ship sophisticated, production-ready AI assistants in minutes.

Key Features

• ReFRAG-inspired micro-chunking, token budgeting, and gate-first filtering for lean, precise context retrieval.
• Performance optimizations including connection pooling, intelligent caching, request deduplication, and async subprocess management.
• One-command setup for a complete retrieval stack with dual SSE/RMCP endpoints, Qdrant, and live reindexing.
• CLI prompt enhancer with multi-pass 'Unicorn mode' for code-grounded and highly intelligent prompt refinement.
• Unified schema and tooling for seamless integration of dense, lexical, and semantic search signals.
• 72 GitHub stars

Use Cases

• Enabling AI platform and IDE tooling teams to quickly establish an MCP-compliant context layer for agents.
• Assisting DevEx and documentation groups in deploying internal assistants that ingest dynamic codebases.
• Solving slow AI agent onboarding and context drift by providing a consistent and automatically updated context stack.