web2bigtable

web2bigtable / Web2BigTable

Public

A Bi-Level Multi-Agent LLM System for Internet-Scale Information Search and Extraction

arxiv.orgabs2604.27221 information-extraction multi-agent-systems self-evolving web-search
23
0
100% credibility
Found May 06, 2026 at 23 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Web2BigTable is a multi-agent AI framework that turns natural language queries into verified structured tables by searching the web through a terminal interface.

How It Works

1
📥 Get it set up

Run one easy command to download and prepare your web research helper.

2
🖥️ Open the screen

Type a simple command to launch the friendly terminal window.

3
💭 Ask for a table

Type a question like 'Make a table of top companies by sales' in the box.

4
👥 Watch the team work

See little helpers search the web together, share notes, and build your table step by step.

5
📋 Review the table

Check the finished table right there, with sources verified.

✅ Copy and use

Copy your perfect table to paste anywhere you need it.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 23 to 23 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 Web2BigTable?

Web2BigTable is a Python-based bi-level multi-agent LLM system that turns natural language queries and target schemas into verified structured tables pulled from internet-scale web sources. It tackles wide searches assembling many entity rows across sites and deep searches chaining multi-hop clues, with an upper orchestrator decomposing tasks and lower workers executing in parallel via shared workspaces. Users get a TUI interface launched by `web2bigtable`, one-click install, and outputs like entity benchmarks or comparative data tables.

Why is it gaining traction?

Its bi-level actor critic for multi-agent coordination and bi-level knowledge transfer stand out, enabling self-evolving decomposition and execution skills without fine-tuning LLMs—benchmarks show it dominating WideSearch (63.5 Row F1, 80.1 Item F1) and hitting 73% on XBench-DeepSearch, beating open-source agents. Developers hook on the parallel workers (up to 10), semantic skill routing, and ops-based execution for reliable web extraction, plus the arXiv-backed research making it a fresh take on multi-task multi-agent reinforcement learning.

Who should use this?

Data scientists scraping web info for tables, like company rankings or event timelines. Analysts building entity databases from heterogeneous sources. Researchers prototyping internet-scale extraction pipelines where single agents fall short on coverage or depth.

Verdict

Worth a spin for web-to-table needs—solid TUI, easy setup, strong benchmarks—but at 23 stars and 1.0% credibility, it's early alpha with risks in production stability. Fork and evolve it yourself; docs and skills ecosystem help bridge the maturity gap.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.