togethercomputer/together-cookbook

What it solves

It provides a comprehensive collection of code examples and guides to help developers implement open-source AI models using the Together AI platform. It bridges the gap between having access to models and knowing how to build complex, production-ready AI workflows.

How it works

The project is organized as a "cookbook" of Jupyter notebooks and code snippets. Developers can copy these recipes to implement specific AI patterns, such as agentic workflows, fine-tuning strategies, and retrieval-augmented generation (RAG) pipelines, all interacting with the Together AI API.

Who it’s for

Developers and AI engineers who want to build applications using open-source models and need practical, runnable examples of advanced LLM techniques.

Highlights

• Agentic Workflows: Examples of serial, parallel, looping, and orchestrator subtask agents, as well as integrations with frameworks like LangGraph, PydanticAI, and DSPy.
• Advanced Fine-tuning: Guides for LoRA, DPO (preference tuning), continual fine-tuning, and specialized tuning for reasoning, function calling, and vision-language models.
• RAG & Search: Implementations of contextual RAG, multimodal RAG, and semantic search with reranking.
• Multimodal Vision: Recipes for OCR, 2D/3D grounding, video understanding, and document parsing using vision-language models.