Data Tooling

Indra OCR trades heavyweight pipelines for a tiny ONNX recognizer, and somehow still beats Tesseract and PaddleOCR on accuracy.

A Rust search engine that auto-indexes folders so AI agents can query code and documents without stuffing their context windows.

A stdlib-only Python tool that pulls years of chat history out of Claude Code, Cursor, Aider, and seven other AI coding assistants into one normalized JSONL format.

OXR turns PDFs and document images into agent-ready Markdown or JSON, and claims #1 on three parsing benchmarks while being small enough to be fast.

It gives AI agents complete social media research workflows—outlier detection, comment mining, competitor teardowns—that produce actionable business artifacts instead of raw API responses.

When several agents each carry their own copy of the docs, you get several versions of the truth — OpenLore is one governed Markdown source they all read from instead.

A research platform that fuses CARLA, AirSim, and OpenSim into one Unreal-based sandbox for training autonomous cars, drones, and embodied humans.

It materializes ontologies and hands you a Lean-checked certificate for every inference, because asking a reasoner to explain itself usually goes nowhere.

WeChat 4.x can merge-forward whole chats as a ZIP, but Claude and Codex ship no share extension to catch it — WeChatBridge registers twelve and routes the archive into your agent, Obsidian, or the clipboard.

Data agents re-infer your business logic from raw tables on every task; EvoOntology builds a versioned semantic layer over your data and evolves it from execution history, publishing only updates that beat their parent.

learn-from-materials turns PDFs, books and papers into interactive learning pages where every claim traces back to a source page.

It scrapes WeChat public articles and serves them as RSS, Markdown, and JSON because Tencent won’t.

Because one person can't be the expert in every industry, AIHOT open-sources its entire automated newsroom — every prompt and threshold — for anyone to retarget at their own field.

Go CLI that replaces guesswork with an interactive picker for model quantizations, branches, and diffusers components.

Open-source Claude skills that run SEO and generative-engine optimization on your actual Google data instead of scraped estimates.

Ask a chat AI to scrape Google Maps and you get a consent wall and zero structured rows; this kit hands Claude a real local scraping engine to drive instead.

An open-source pipeline that curates, licenses, and retrieves task-specific instructions so agents don't have to wing it.

It builds an agent-ready knowledge base over your local files without dragging them into a proprietary workspace.

Utopia ingests your documents into a self-hosted, time-aware knowledge graph so you can ask what was true last quarter—and prove it with citations.

VBench decomposes video generation quality into fine-grained, measurable dimensions with human-aligned automated metrics, letting developers evaluate everything from pixel fidelity to physics realism.