Foundation Guides

The Foundations guides introduce the core building blocks of Beluga AI v2. Each guide builds on the previous one, starting with the agent abstraction that ties everything together, then drilling into the LLM layer that powers reasoning, and finally covering the supporting systems for structured data extraction and prompt management. These guides assume familiarity with Go but no prior experience with AI frameworks. By the end, you will understand Beluga’s key design patterns — the registry pattern for pluggable providers, iter.Seq2 for streaming, functional options for configuration, and middleware for composable behavior — and how they work together to form a production-ready agentic system.

We recommend reading the guides in order:

Guide	Description
Building Your First Agent	Create a complete AI agent with tools, streaming, and reasoning. Introduces the agent runtime, the ReAct loop, and handoffs between specialized agents.
Working with LLMs	Configure language models through the unified ChatModel interface. Covers provider setup, streaming with iter.Seq2, middleware composition, hooks, and multi-provider routing.
Structured Output	Extract typed Go structs from LLM responses. Covers schema generation, retry strategies, classification patterns, and production validation.
Prompt Engineering	Design, template, version, and optimize prompts. Covers the PromptManager, Builder for cache-optimal ordering, few-shot patterns, and A/B testing.