ChatGH

chopratejas / headroom

A local-first context compression layer for AI agents. Compresses tool outputs, logs, RAG chunks, files, and conversation history by 60–95% before sending to the LLM — with reversible retrieval via CCR. Works as a Python/TS library, zero-code proxy, MCP server, or one-command agent wrapper for Claude, Codex, Cursor, and more.

microsoft / markitdown

Microsoft's lightweight Python tool that converts PDF, Word, Excel, PowerPoint, images, audio, HTML, and more into Markdown — preserving headings, tables, and links. Designed for LLM and RAG pipelines, with CLI/Python API, MCP server for AI agents, plugin extensibility, and optional Azure cloud and LLM-powered OCR.

affaan-m / ECC

A cross-harness AI coding agent optimization system with 60 specialized agents, 232 skills, automated hooks, security scanning, continuous learning, and memory persistence. Works across Claude Code, Codex, Cursor, OpenCode, Copilot and more, covering 12+ language ecosystems for production-ready agentic development.

D4Vinci / Scrapling

Scrapling is an adaptive Python web scraping framework that handles everything from single requests to full-scale crawls. Its smart parser automatically relocates elements after website changes, built-in fetchers bypass anti-bot systems like Cloudflare, and the spider framework supports concurrent crawling with pause/resume, proxy rotation, and AI integration via MCP server.

nesquena / hermes-webui

Self-hosted web interface for Hermes Agent with full CLI parity. Three-panel layout with session management, streaming chat, and workspace file browser. Features persistent memory, self-improving skills, cron scheduling, multi-provider support (OpenAI/Anthropic/Google/DeepSeek/local models), 10+ messaging platforms, password/passkey auth, 11 skins, and Docker deployment. Pure Python + vanilla JS, no build step.

reconurge / flowsint

An open-source OSINT graph investigation platform with visual node exploration and 30+ automated enrichers across domains, IPs, crypto wallets, social profiles, and more. Self-hosted, privacy-first, with Neo4j graph engine and N8n workflow integration for cybersecurity analysts and investigators.

OpenBMB / VoxCPM

A tokenizer-free text-to-speech system using diffusion autoregressive architecture. The 2B-parameter VoxCPM2 supports 30 languages, voice design from text descriptions, controllable voice cloning with style guidance, and 48kHz studio-quality audio output. Fully open-source under Apache-2.0, with real-time streaming and production deployment via vLLM-Omni.

stefan-jansen / machine-learning-for-trading

Comprehensive ML for algorithmic trading codebase with 150+ notebooks. Covers end-to-end workflow from market/fundamental/alternative data sourcing, alpha factor engineering, to strategy backtesting. Models range from linear regression, random forests, gradient boosting to CNN, RNN, autoencoders, GANs, and deep reinforcement learning. Includes NLP pipelines for SEC filings, earnings calls, and news sentiment analysis.

jamwithai / production-agentic-rag-course

A 7-week hands-on course to build production-grade Agentic RAG systems from scratch. Covers infrastructure setup, arXiv data ingestion, BM25 keyword search, hybrid retrieval, LLM integration with streaming, Langfuse monitoring, Redis caching, LangGraph agent orchestration, and Telegram bot — using FastAPI, OpenSearch, Airflow, and Docker.