← All comparisons
LemonCrow vs cocoindex-code

What cocoindex-code says, vs. what it scored.

AST-based semantic code search built on CocoIndex, a Rust data-transformation engine; tree-sitter parsing plus embeddings.

What cocoindex-code says about itself
“A lightweight, effective (AST-based) semantic code search tool for your codebase... Instant token saving by 70%.”
Publishes some numbers ...never against another search tool ~1.6k stars ↗
What it actually scored — same 14 repos, same 7,213 queries as every other tool
Tool MRR p95 p100
LemonCrow +semantic (BGE) 0.727 390ms 1057ms
LemonCrow lexical (default) 0.676 134ms 319ms
cocoindex-code 0.557 595ms 2061ms

70% token-saving figure is real. Strongest challenger here -- 0.557 MRR vs. LemonCrow's 0.727 (semantic) / 0.676 (lexical).

By query kind -- same benchmark, broken out (no reps in this eval: one deterministic pass per query)
Kind LemonCrow +semantic LemonCrow lexical cocoindex-code
definition 0.873 (n=1570) 0.871 (n=1570) 0.648(n=1570)
content 0.873 (n=1444) 0.864 (n=1444) 0.696(n=1444)
semantic 0.759 (n=1800) 0.576 (n=1800) 0.590(n=1800)
swebench 0.500 (n=1908) 0.493 (n=1908) 0.371(n=1908)
sessions 0.587 (n=491) 0.571 (n=491) 0.455(n=491)

n = query/gold pairs of that kind, out of 7,213 total -- every provider scored on all 5 kinds.

By repo -- all 15 repos in the corpus, same query set
Repo LemonCrow +semantic LemonCrow lexical cocoindex-code
astropy/astropy 0.772 0.715 0.614
atelier-ws/atelier-dev 0.467 0.477 0.356
atelier/atelier 0.594 0.557 0.457
django/django 0.689 0.652 0.511
matplotlib/matplotlib 0.801 0.747 0.631
mwaskom/seaborn 0.814 0.768 0.711
pallets/flask 0.735 0.671 0.518
psf/requests 0.840 0.803 0.691
pydata/xarray 0.815 0.764 0.660
pylint-dev/pylint 0.856 0.784 0.706
pytest-dev/pytest 0.826 0.739 0.625
scikit-learn/scikit-learn 0.740 0.669 0.513
sphinx-doc/sphinx 0.637 0.580 0.449
sympy/sympy 0.694 0.637 0.504
torvalds/linux 0.726 0.668 0.500

MRR per repo: n-weighted blend across all 5 query kinds, same 7,213-query run.

The true story

Same 14 repositories, same 7,213 query/gold pairs as every tool here, cocoindex-code included. Full methodology, every raw number, and the other 9 tools →