ColGrep pairs code-search filters with semantic matching
Weaviate Podcast says LightOn built models small enough to index a codebase locally on a laptop, and credits ColGrep with fewer repeat queries and fewer tokens spent on search.
TLDR
ColGrep combines regex and file filters with semantic matching—searching by meaning—according to Weaviate Podcast. The podcast says LightOn trained models small enough to index a codebase locally on a laptop, including a 70-million-parameter code retriever it says outperformed everything up to 300 million parameters. It credits the ColGrep approach with fewer repeat queries, fewer tokens spent on search and better answers on the hardest queries.
Combined views
21
1 Source, first seen 15d ago