I spent my most of day's on a new team grepping through 80,000 lines of code trying to find where authentication worked.
Four hours. Three teammates interrupted. Twelve dead ends across files I didn't understand. The code was fine — it was well-written, well-organized, reasonably documented. The tooling was the problem. I was using grep to understand something that wasn't a text search problem. Code has structure: call edges, import chains, type hierarchies, AST relationships. Grep ignores all of it.
That day stuck with me. I kept running into the same pattern on different teams, different codebases, different languages. Every time I joined something new or came back to a project after six months away, the first few days were archaeology. Tracing calls manually. Reconstructing context that should have been queryable.
I built trelix to fix this. It's an open-source code intelligence engine that indexes any repository with Tree-sitter, embeds every symbol, and answers natural-language questions using hybrid BM25 + vector + call-graph search. It works offline. No API key needed. Zero infrastructure.
pip install "trelix[local]"
