Choose a case file
Start by category, framework, difficulty, or the kind of application you want to build.
Open-source intelligence / File 050
Clone.
Learn.Build.
A curated field guide to 50 open-source GenAI applications and Hugging Face Spaces—selected for cloneability, setup clarity, architecture, and real learning value.
Curated evidence
50Curated projects
09Focused guides
40%Cloneability weight
2Open data formats
Evidence board / Categories
Follow the trail that fits your build.
Skip the random link dump. Start with a focused collection based on the architecture, framework, or workflow you actually want to understand.
CASE / 01
RAG Projects
PDF chatbots, vector search, document assistants, and retrieval pipelines.
CASE / 02Agentic AI
Tool-using agents, agentic retrieval, routing, and workflow automation.
CASE / 03LangGraph
Graph-based agent flows, explicit state, branches, and retry loops.
CASE / 04Multi-Agent
Crew workflows, collaborating roles, delegation, and shared state.
CASE / 05MCP Projects
Model Context Protocol servers, reusable tools, and app wrappers.
CASE / 06Beginner Friendly
Smaller projects with clearer entry points and faster setup paths.
Selection protocol
Usefulness before popularity.
Projects are evaluated as learning and building references. A famous demo with missing source or unclear setup should not outrank a project developers can actually inspect and adapt.
Cloneability
40%
Documentation
20%
Architecture
15%
Learning value
15%
Popularity
10%
Field procedure
From discovery to working code.
The Starter Pack is a catalog and dataset—not one giant boilerplate. Use it to find a project whose shape matches what you want to build.
Inspect the evidence
Compare the stack, setup estimate, architecture, source availability, and learning notes.
Clone the source
Most Hugging Face Spaces are Git repositories. Clone one directly and inspect its entry point.
Adapt the pattern
Reuse the retrieval, agent, tool, UI, or workflow pattern as a foundation for your own project.
The investigation starts here