Shubhamsaboo/awesome-llm-apps

A growing collection of 100+ runnable AI agent and RAG application examples — clone, customize, ship.

What it is

A Python-based collection of working LLM application examples — agents, RAG pipelines, chatbots, multi-modal apps, voice assistants — each as a standalone directory with the code, requirements, and a brief README explaining what it does and how to run it. Aimed at engineers who want concrete examples to learn from or fork into their own projects. Curated by Shubham Saboo with TheUnwindAI as the editorial brand.

Key features

100+ working examples across categories: AI agents, RAG, multi-modal, voice, code-assistants, customer-support.
Each example is self-contained — clone the directory, install requirements, run.
Coverage spans many provider/framework combinations (OpenAI, Anthropic, local models, LangChain, LlamaIndex, CrewAI, etc.).
Apache 2.0 licensed.
Newsletter / content brand TheUnwindAI funds ongoing curation.

Tech stack

Python primary.
Per-example dependencies vary; each ships its own requirements.txt.
Streamlit / Gradio are popular UI choices among the examples.

When to reach for it

You're learning LLM app patterns by reading working code.
You're prototyping a specific kind of app (RAG over PDFs, voice agent, etc.) and want a starting point to fork.
You're a content creator citing concrete examples in tutorials.

When not to reach for it

You want production-grade reference architectures — examples are tutorial-flavored, not battle-tested.
You want a single coherent codebase — each example is independent; there's no shared library to reuse across them.
You want a license-clean training corpus that excludes content marketing — TheUnwindAI's curation includes promotional framing.

Maturity signal

112k stars, 16k forks, Apache 2.0, last push 2026-06-01. 2-year-old project that became a popular "show me working code" reference as the LLM-app space exploded. Open-issues count of 11 is exceptionally low — the maintainer triages aggressively, partially because content quality is the brand's value proposition. Apache 2.0 + brand-backed curation make this safe to fork into commercial work.

Alternatives

LangChain templates (langchain-ai/langchain repo) — use for first-party LangChain reference apps.
gpt-engineer-org/gpt-engineer — use for end-to-end "describe an app, get the code" examples.
Provider-specific cookbooks (OpenAI Cookbook, Anthropic Cookbook) — use when you want vendor-curated examples.

Notes

The "100+ apps" claim grows over time as new examples land; check the directory count rather than the README text for the current total. Apache 2.0 license is the safer choice among popular LLM-app collections. The TheUnwindAI brand attachment is editorial — content references the newsletter — but doesn't carry usage restrictions.

Wiki: Shubhamsaboo/awesome-llm-apps