Metadata-Version: 2.4
Name: quering
Version: 0.1.0
Summary: A local UI package for querying pooled embedding zips against original markdown chunks
Author: Wenxi Wang
License-Expression: MIT
Project-URL: Homepage, https://example.invalid/quering
Requires-Python: >=3.10
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: streamlit>=1.36
Requires-Dist: sentence-transformers>=2.6.0
Requires-Dist: numpy>=1.24
Dynamic: license-file

# quering

A local UI package for querying pooled embedding zips against the original markdown chunk zip.

## What it does

- launches with the `quering` command
- accepts a pooled embedding zip from the previous `poolin` step
- accepts the original markdown chunk zip from the `MarkdownHeaderTextSplitter` step
- lets you choose an embedding model for the query sentence
- ranks the highest scoring pooled groups
- maps the best matches back to the original markdown chunk texts
- exports a zip with results, matched chunks, and an LLM-ready prompt

## Install

```bash
pip install quering
```

## Run

```bash
quering
```

## Output zip contents

- `querying_summary.json`
- `querying_results.csv`
- `llm_prompt.txt`
- `inputs_used.json`
- `*_results.jsonl` (optional)
- `*_results.npz` (optional)
- `matched_chunks/` (optional)

## Notes

- Best results come from using the **same embedding model** for the query that was used to create the pooled embeddings.
- This package is an original implementation for local experimentation.
