xiaoqi-7

xiaoqi-7 / SheetasToken

Public

Implementation and resources for Sheet as Token, a graph-enhanced framework for multi-sheet spreadsheet understanding and retrieval.

19
6
100% credibility
Found May 14, 2026 at 20 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

An academic codebase implementing a two-stage AI pipeline for encoding spreadsheets and retrieving relevant sheets from multi-sheet Excel files based on natural language queries, with training scripts and an optional web service for running searches.

How It Works

1
🔍 Discover SheetAgent

You learn about a helpful tool that uses smart thinking to find the right sheet in big Excel files when you ask a question in plain words.

2
💾 Get the files

Download the ready-to-use project files to your computer so you can start helping with your spreadsheets.

3
🧠 Add smart brains

Point the tool to pre-made thinking models trained on spreadsheets, like telling it where the knowledge lives.

4
🚀 Launch the finder

Start the friendly service with a simple go, and it wakes up ready to scan your files.

5
📤 Send files and question

Share links to your Excel workbooks and type your question, like 'show sales by region'.

6
⏳ Let it search

The tool quietly reads your files and picks the best matching sheets while you wait.

✅ Get perfect sheets

Receive a list of the most relevant sheets, so you jump straight to the data you need without hunting.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 20 to 19 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 SheetasToken?

SheetasToken is a Python framework for retrieving relevant sheets from multi-sheet spreadsheets using graph-enhanced representations. Feed it Excel URLs and a natural language query via its FastAPI endpoint, and it returns top-ranked sheets by name and index—ideal for agents querying workbook data without scanning everything. Built on Transformers and Torch, it provides training scripts and a two-stage pipeline as implementation resources for custom setups.

Why is it gaining traction?

Unlike basic keyword search, this graph-enhanced framework uses sheet tokens with relational priors (schema, shape) for smarter cross-sheet ranking, beating naive bi-encoders in retrieval accuracy. Devs grab it for the plug-and-play API handling async jobs and the full reproduction of arXiv paper results, similar to alphazero implementation github or yolo implementation github projects. Zero-config serving on HTTPS URLs hooks RAG builders fast.

Who should use this?

ML engineers tuning retrieval for spreadsheet-heavy domains like finance or analytics, where queries span workbook sheets. Researchers validating table QA benchmarks or extending graph-enhanced frameworks with their data. Teams needing tod implementation resources and tools for query-sheet matching in LLM pipelines.

Verdict

Grab it for research reproduction or prototyping graph-enhanced retrieval—docs cover setup, training, and API clearly. But 19 stars and 1.0% credibility score signal early maturity; expect tweaks for production scale.

(187 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.