Metadata-Version: 2.4
Name: lensify
Version: 0.2.0
Summary: Lensify upgraded local semantic search
Home-page: https://github.com/wakeups1dd/lensify
Author: Siddhant Gangawane
Author-email: siddhantgangawane335@gmail.com
License: MIT
Classifier: Programming Language :: Python :: 3
Classifier: License :: OSI Approved :: MIT License
Classifier: Operating System :: OS Independent
Requires-Python: >=3.8
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: sentence-transformers; extra == "embeddings"
Requires-Dist: numpy
Requires-Dist: faiss-cpu
Requires-Dist: PyPDF2
Requires-Dist: scikit-learn
Requires-Dist: nltk
Requires-Dist: tqdm
Dynamic: license-file
Dynamic: summary

# Lensify — Upgraded Local Semantic Search Engine (drop-in ready)

This upgraded Lensify prototype includes the following accuracy improvements:
- FAISS HNSW index for fast, high-recall retrieval
- Support for high-quality embedding models (BAAI/bge-*/sentence-transformers)
- Cross-Encoder re-ranker (ms-marco or similar) for final ranking
- Sentence-boundary chunking with overlap
- Text normalization and query expansion
- New CLI commands: doctor, stats, rebuild, query, export

IMPORTANT:
- Large models (BGE large, cross-encoders) are optional but recommended.
- To get best results install dependencies:
  pip install -r requirements.txt

Quickstart:
1. Install dependencies (recommended): pip install -r requirements.txt
2. Rebuild index: lensify rebuild /path/to/docs
3. Query: lensify query /path/to/docs "payment risk simulation" --k 5
4. Use doctor to check environment: lensify doctor

Notes:
- This is a prototype. For production usage ensure you have enough RAM and disk for large models.
