Paper2Code is a research system that automatically generates runnable code repositories from machine learning research papers. The core pipeline, PaperCoder, uses a multi-agent LLM architecture (planning→analysis→coding) implemented as a sequence of Python scripts orchestrated via shell commands. It accepts papers in PDF (converted to JSON via s2orc-doc2json), JSON, or LaTeX format, and produces a complete Python project with config, model definitions, training loops, and evaluation scripts. The codebase supports both OpenAI API (o3-mini by default) and open-source models via vLLM (e.g., DeepSeek-Coder-V2-Lite-Instruct). It includes a debug stage that iteratively fixes runtime errors using LLM-generated SEARCH/REPLACE patches, and an evaluation module that scores generated repositories (reference-free or reference-based) using LLM judgment. The repository also bundles the Paper2Code benchmark dataset (90 papers from top ML conferences with gold repositories) and prompts for evaluation. Although primarily an application, it has research artifacts (evaluation scripts, prompt templates, benchmark data) that give it a hybrid character.
This is the umbrella monorepo for Evo CRM Community, an open-source AI-powered customer support platform. The repository itself does not contain application source code; it orchestrates seven independent microservices (auth, CRM, frontend, AI processor, core service, bot runtime, and event flow) via Git submodules and Docker Compose/Swarm configurations. The primary function of this repo is to provide a single entry point for cloning, building, seeding, and deploying the full stack. It includes a nginx API gateway for path-based routing to backend services, CI/CD workflows for building and publishing Docker images, migration version collision guards, and extensive release documentation. The actual business logic, AI agent execution, conversation management, and user interface reside in the submodules.
This repository is an automated IPTV playlist aggregation pipeline. It operates as a daily GitHub Actions workflow that fetches live TV channel URLs from dozens of remote sources, parses them (M3U and plain text), classifies by channel category using hand-maintained dictionary files (e.g., CCTV, provincial channels, sports, movies), applies name normalization via OpenCC, deduplicates, sorts, and outputs merged TXT and M3U files. It also includes a stream validation subsystem (blacklist1.py, blacklist2.py) that checks URL health, records failures with timestamps, and maintains a persistent failure cache and host reputation tracker via a JavaScript state module. Historical snapshots and output archives are stored for rollback. The system is heavily customized for Chinese IPTV sources and lacks automated tests, but the core pipeline pattern and the stream validation state management are reusable.