haid is a "hearing aid" for agentic coding environments such as Claude Code, Codex, and Gemini. It provides a local, high-performance search and retrieval interface designed specifically for LLM-based agents to consume.
By embedding sift as its core engine, haid allows agents to:
- Search local codebases with intent-driven hybrid retrieval.
- Recommend relevant files based on current coding context.
- Extract concise, agent-ready information from search hits.
- Single Rust Binary: No external database, daemon, or long-running service.
- Sift-Powered: High-performance SIMD-accelerated scoring and hybrid strategy (Lexical + Semantic).
- Agent-Ready Output: Structured JSON or formatted text optimized for LLM token efficiency.
- Local Sovereignty: All extraction, search, and inference happen locally.
curl --proto '=https' --tlsv1.2 -LsSf https://github.com/rupurt/haid/releases/latest/download/haid-installer.sh | shDownload the latest pre-built binaries and installers for your platform from the GitHub Releases page.
# Search the current directory
haid search "how do I handle authentication?"
# Recommend files related to a specific context
haid recommend src/auth.rs
# Extract information from a specific file with context
haid extract src/main.rs "show the entry point"- AGENTS.md: Shared guidance for AI agents working on this project.
- CONSTITUTION.md: Repository design and operational principles.
- ARCHITECTURE.md: Deep dive into the hexagonal engine and component boundaries.
- INSTRUCTIONS.md: Formal workflow and operational guidance (Keel/Missions).
- CONFIGURATION.md: Guide to
haid.tomland environment variables. - EVALUATIONS.md: How to measure retrieval quality and extraction accuracy.
- RELEASE.md: The automated release and distribution process.
This project is licensed under the MIT License.