Discover, learn, and launch AI projects across machine learning, AI agents, generative AI, automation, chatbots, and more.
The framework most LLM applications are still assembled from.
Run open models on your own machine with one command.
LLM inference in plain C/C++, fast enough to run on a laptop.
High-throughput serving built around PagedAttention.
A polished self-hosted chat front-end for local and remote models.
A visual platform for building and operating LLM apps.
The project that made autonomous agents a mainstream idea.
An agent that writes code, runs it, and fixes what breaks.
The browser interface that made local image generation usable.
A node graph that exposes the whole diffusion pipeline.
Speech recognition that holds up on accents, noise and 90+ languages.
Fair-code workflow automation with AI steps built in.
Twelve build areas, from chatbots and agents to vector search and fine-tuning.
Retrieval-augmented generation — grounding a model in your own documents and data.
5 projectsThe scaffolding around a model: UI builders, experiment tracking and app frameworks.
5 projectsRunning models fast and cheaply: local runtimes, inference servers and OpenAI-compatible APIs.
4 projectsEmbedding stores and similarity search — the memory layer under most RAG systems.
3 projectsConversational apps you can actually ship — chat UIs, assistants and the plumbing behind them.
2 projectsProjects where the model plans, calls tools and works a task through to a result on its own.
2 projectsDiffusion pipelines, node graphs and browser front-ends for generating and editing images.
2 projectsSpeech recognition, text-to-speech and voice cloning you can run yourself.
2 projectsDetection, segmentation and tracking — the frameworks real vision products are built on.
2 projectsAdapting an existing model to your data — LoRA, full fine-tunes and the tooling around them.
2 projectsWiring models into multi-step processes that run without a human in the loop.
2 projectsSeveral specialised agents co-ordinating on one job — roles, hand-offs and shared state.
2 projectsEvery entry is open source, currently maintained, and documented well enough to start from.
Each entry links straight to its repository, so you can read the code and judge the thing before you commit an afternoon to it.
Languages, frameworks and infrastructure are listed per project — not left as "some AI" — so you know what you are signing up to run.
Beginner through expert, set against what the project actually asks of you: the toolchain, the hardware and the prerequisites.
MIT, Apache, AGPL or something bespoke — stated on the card, because it decides whether you can ship what you build.
Filter to what runs on your own hardware, or to what only needs a key. Both are legitimate; the catalogue does not assume.
Category, technology and tag hubs cross-cut the same catalogue, so you can arrive from whichever direction you already think in.
