v0.1.8

Friedrich-M/UniMatev0.1.8Mar 23, 2026by chonknick

Upgrading?

AI Summary

Overhauls reranking performance by 3.3x using a single rerank pass, optimized ONNX model format (ORT), memory-mapped file loading, and multi-threaded inference.

Key Highlights

  • 3.3x faster reranking performance
  • Single rerank pass across all packages
  • Introduction of `.ort` model format with mmap loading
  • Multi-threaded inference support (half available cores)

New Features

  • Reranking performance overhaul (ORT format, mmap, multi-threading)

Full Release Notes

## What's Changed

### Reranking Performance Overhaul (3.3x faster)

- **Single rerank pass** — collect all FTS5 results across packages, rerank once instead of per-package (eliminates redundant ONNX session loads)
- **ORT format** — pre-optimized flatbuffer model (`reranker.ort`) for faster deserialization, with automatic fallback to `.onnx`
- **mmap model loading** — zero-copy memory-mapped file access via `commit_from_memory_directly`
- **Multi-threaded inference** — uses half available cores (capped at 4) for ONNX intra-op parallelism

### Benchmarks

| Scenario | v0.1.7 | v0.1.8 | Speedup |
|----------|--------|--------|---------|
| All packages (11) | 1.01s | **0.31s** | 3.3x |
| Single package | 0.09s | **0.04s** | 2.3x |

Related Repositories