53.5K ·GitHub
What is this?
Anthropic's collection of notebooks and recipes showing fun and effective ways of using Claude, with copy-able code snippets you can drop straight into your own projects.
Anthropic's collection of notebooks and recipes showing fun and effective ways of using Claude, with copy-able code snippets you can drop straight into your own projects.Is it one course to follow start to finish, or separate recipes I can dip into?It is a menu of standalone recipes grouped into five sections: Capabilities, Tool Use and Integration, Third-Party Integrations, Multimodal Capabilities, and Advanced Techniques.
It is a menu of standalone recipes grouped into five sections: Capabilities, Tool Use and Integration, Third-Party Integrations, Multimodal Capabilities, and Advanced Techniques.How do I actually run one of these recipes?Pick a recipe from the table, open its notebook in a Jupyter environment, and run the cells with your Claude API key set.
Pick a recipe from the table, open its notebook in a Jupyter environment, and run the cells with your Claude API key set.Will I need an API key, or can I just read along?You'll need a Claude API key to make the most of the examples; signing up for one is free.
You'll need a Claude API key to make the most of the examples; signing up for one is free.Do I have to write Python to follow along?The notebooks are written primarily in Python, though the concepts can be adapted to any language that can interact with the Claude API.
The notebooks are written primarily in Python, though the concepts can be adapted to any language that can interact with the Claude API.I have never touched the Claude API before — is there a gentler starting point?Yes, the maintainers point API newcomers to the Claude API Fundamentals course first to build a foundation before diving into the recipes.
Yes, the maintainers point API newcomers to the Claude API Fundamentals course first to build a foundation before diving into the recipes.What kinds of tasks do the basic recipes cover?Text and data classification, retrieval augmented generation for grounding answers in external knowledge, and text summarization.
Text and data classification, retrieval augmented generation for grounding answers in external knowledge, and text summarization.What multimodal tricks does it teach?Getting started with images, vision best practices, interpreting charts and graphs, extracting content from forms, and image generation by pairing Claude with Stable Diffusion.
Getting started with images, vision best practices, interpreting charts and graphs, extracting content from forms, and image generation by pairing Claude with Stable Diffusion.Which third-party services show up in the recipes?Pinecone vector databases and Voyage AI embeddings for retrieval, plus Wikipedia and web pages as external knowledge sources.
Pinecone vector databases and Voyage AI embeddings for retrieval, plus Wikipedia and web pages as external knowledge sources.What does the advanced techniques section cover?Using Haiku as a sub-agent with Opus, passing PDFs to Claude as text, automated prompt evaluations, JSON mode, a moderation filter, prompt caching, and a cost optimization checklist.
Using Haiku as a sub-agent with Opus, passing PDFs to Claude as text, automated prompt evaluations, JSON mode, a moderation filter, prompt caching, and a cost optimization checklist.What are similar projects?courses, llama-cookbook, ai-engineer-notebooks, Hands-On-Large-Language-Models, llm-cookbook, agentic-rag-for-dummies, cookbook, openai-cookbook
courses, llama-cookbook, ai-engineer-notebooks, Hands-On-Large-Language-Models, llm-cookbook, agentic-rag-for-dummies, cookbook, openai-cookbookWhich topics is it in?cookbook, claude-api, notebooks, snippets, vision, tutorials, jupyter-notebook, examples, embeddings, jupyter notebook, multimodal, prompt-engineering
cookbook, claude-api, notebooks, snippets, vision, tutorials, jupyter-notebook, examples, embeddings, jupyter notebook, multimodal, prompt-engineering