I Built trelix Because I Was Tired of Grepping My Way Through Codebases

Short summary

Author built trelix, an open-source code intelligence engine that indexes repositories with Tree-sitter and answers natural-language questions about code structure using hybrid search (BM25 + vectors + call graphs). Trelix defaults to a single SQLite file with zero external infrastructure, with optional scaling to LanceDB or Qdrant for larger repositories. The 3-tier adaptive router classifies queries and applies appropriate retrieval strategies for symbol lookup, feature flows, and dependency analysis.

• •Open-source trelix tool uses Tree-sitter, hybrid search (BM25 + vectors + call graphs) to understand codebases via natural language
• •Single SQLite index by default with zero infrastructure; optional scaling to LanceDB (100k+ chunks) or Qdrant (500k+ chunks)
• •3-tier router adapts retrieval strategy based on query intent: direct answers, 8-intent classification, or multi-step decomposition

Generated with AI, which can make mistakes.