finds.dev

openai/openai-cookbook

Examples and guides for using the OpenAI API

View on GitHubHomepage ↗

★ 74,141Jupyter NotebookMITupdated Jun 2026

// the take

The official OpenAI recipe book: a large collection of Jupyter notebooks showing how to use the OpenAI API for embeddings, fine-tuning, function calling, agents, RAG, and more. It's maintained by OpenAI staff and updated alongside new API features, so it tracks current capabilities reasonably well. The target audience is developers starting out with the API or looking for working reference code.

Breadth is genuinely useful — covers embeddings, RAG, fine-tuning, structured outputs, function calling, and the newer Agents SDK in one place. Notebooks are runnable as-is with just an API key, which lowers the friction from 'curious' to 'working'. The agents_sdk examples (parallel agents, memory compaction, evaluate_agents) reflect real patterns you'd actually use in production rather than toy demos. Active maintenance — last push was yesterday, so examples don't rot against API changes the way most third-party tutorials do.

Everything is tightly coupled to OpenAI's own products — the 'migrate-from-claude-agent-sdk' notebook is a tell; this is partly a marketing asset, not just an educational one. Notebook sprawl is real: 80+ examples with no clear learning path, and quality varies widely between them. Error handling and production concerns are almost entirely absent — notebooks show the happy path, which teaches bad habits. The Azure section is a graveyard of archived notebooks that contradict the live ones, signaling that cross-cloud content gets neglected as soon as a new API version ships.

We read the README, the directory tree and a key source file in June 2026. Nobody pays to be listed. How we judge →

// at a glance

Stars
74,141
Forks
12,548
Language
Jupyter Notebook
License
MIT
Last commit
12 Jun 2026

// nearby

If openai-cookbook isn't quite it.

All repos →