Grip – Retrieval engine, no embeddings, no vector DB
1 points
1 hour ago
| 1 comment
| pypi.org
| HN
Nivens_McTwisp
1 hour ago
[-]
Retrieval engine. Indexes text, finds what you're looking for. No embeddings. No vector database. No GPU. No learned parameters. Words are positioned in a graph and neighbors vote on relevance. pip install getgrip grip ingest --source ./your-code grip search "authentication handler" Free up to 10K chunks. No signup. No API key. Works offline. Benchmarks (21,708 queries, 6 BEIR datasets): 0.464 NDCG@10 core engine, ~0.58 with optional MiniLM reranker. Beats BM25 on all 6. Logs with timestamps in the repo. Confidence scoring (HIGH/MEDIUM/LOW/NONE) so your LLM knows when to say "I don't know" instead of hallucinating. getgrip.dev | GitHub: github.com/Grip-Hub/getgrip.dev
reply