Razshy

Razshy / Resonance-Search

Public

A new search paradigm where documents have gravity, queries converge into basins, and multi-signal scoring uses interference instead of linear fusion.

29
1
100% credibility
Found Apr 04, 2026 at 29 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Rust
AI Summary

Resonance Search is a fast Rust library for document retrieval that models embeddings as gravitational fields and uses convergence and interference scoring for superior accuracy without query-time machine learning.

How It Works

1
📖 Discover smarter search

You find Resonance Search, a tool that makes finding info in your documents feel like magic, pulling the best matches together like gravity.

2
🧪 Try the quick demo

Run the simple example with sample legal files to see it instantly rank the most relevant chunks better than regular searches.

3
📁 Add your own files

Load your documents or texts, and it automatically weighs them by how connected they are, ready for smart searching.

4
🔍 Ask a question

Type your search words, maybe add a special number or code from your query, and watch it explore.

5
⚡ See results converge

Your query gently moves toward the deepest group of perfect matches, delivering top results in milliseconds that beat ordinary methods.

6
🚀 Launch for everyone

Turn it into a web search service or embed in your tool, handling thousands of documents lightning-fast.

✅ Perfect searches every time

Now you always find the exact info you need, with rankings that understand context and connections like never before.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 29 to 29 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is Resonance-Search?

Resonance Search is a Rust-based hybrid search engine that combines vector embeddings, text matching, and regex patterns into a single score, using gravitational "basins" to pull queries toward relevant document clusters. Unlike standard engines that just sum similarities, it runs quick gradient descent steps so queries adapt and converge on the best results, delivering sub-millisecond latency with no ML models at query time. Developers get a simple API to index chunked docs with precomputed embeddings, search via text/embedding/pattern, and run a REST server or Docker container out of the box.

Why is it gaining traction?

It crushes BM25, pure vector search, and linear fusions in benchmarks like CUAD legal contracts and BEIR datasets, with NDCG@10 gains up to 63% over baselines and improvements scaling with corpus size. The interference scoring only boosts when signals align, avoiding noise, while convergence uncovers hidden clusters that one-shot retrieval misses. Rust ensures it's blazing fast and embeddable, with built-in benches, weight optimizers, and eval metrics for easy testing.

Who should use this?

RAG builders needing robust multi-signal ranking for codebases, legal docs, or enterprise search where vector+text alone falls short. Teams evaluating paradigm search group alternatives or resonance search for anomaly detection, geo queries, or recruitment tools will like the pattern support and github code/files/history search capabilities. It's ideal for resonance talent search exam prototypes or vibration test logs where precise basin-finding matters.

Verdict

Promising early experiment with strong benchmarks and thorough docs/tests, but at 29 stars and 1.0% credibility, it's not production-ready yet—prototype it for resonance search test cases first. Grab the commercial license if AGPL copyleft bites.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.