Eulix: open-source code indexing that only calls an LLM when reasoning is actually needed
blune-foo · reddit · 2026-10-15
The author of Eulix, an open-source local indexing tool for large codebases, explains its core design: the LLM is the last resort, not the first call.
- Rust + tree-sitter parsing into symbols, call graphs and embeddings; "where is foo" is an exact lookup, "who calls foo" a graph walk — neither touches a model
- Open-ended queries go through hybrid retrieval (exact symbols, BM25, IVF vectors, call-graph expansion, subsystem signals, test files demoted, MmrSelect ranking); only questions needing reasoning reach an LLM via local Ollama or a remote provider
- Linux kernel benchmark: 37.3M lines across 64,460 files parsed in 57s on 12 threads, producing 773K nodes / 1.5M edges; embedding 837K chunks took 27 min on a single RX 6700 XT; parser peak memory cut from 8GB to 3.5GB
- Ships a beta MCP server (stdio) with query, querywithcot and setrepopath tools, keeping the knowledge base warm and adapting retrieval to known agent symbols
- Known limits: approximate relationship resolution (dynamic dispatch, reflection), unexpanded C/C++ macros leave half of kernel files partially parsed
Apache-2.0, no account or API key needed.
More from coding & agent
- Agent Teams Cost Up to 5.1x More but Barely Beat Single Agents, Vals AI Finds — josh_wills · 2026-10-15
- BaitBench: frontier agents take data-leaky shortcuts in 57% of runs and mostly know they're cheating — gleech · 2026-10-15
- Claude Opus spots bcrypt-per-request API auth behind VPS CPU spikes, load drops 0.9 to 0.4 — DanielLockyer · 2026-10-15
- Burkov: labs are now training coding LLMs on git history, exactly as he predicted 9 months ago — burkov · 2026-10-15
- Cohere research: swarms of small models reach frontier-level software engineering — Cohere_Labs · 2026-10-15
- 12 GitHub repos to supercharge Claude Code: context, parallel sessions, automation — Roger_M_Taylor · 2026-10-15