11. Building LLM Agents with uni-agent

Coming soon

This chapter is an outline. The full draft ships in a follow-up PR.

What this chapter will cover

uni-agent is Uni's LLM agent framework: a small set of abstractions for chat sessions, tool calling, and model orchestration. This chapter builds a working agent from the ground up.

Concepts introduced:

• What an "agent" is in this framework, as opposed to a raw chat completion.
• ChatSession for turn-taking conversations with persistent state.
• Registering Scala functions as tools the model can call.
• Switching between providers (AWS Bedrock, etc.) without changing your business logic.
• Testing agent code — what is deterministic, what is not, and how to assert on either.

Reference you can read now

• Agent Framework — Overview
• LLM Agent
• Chat Sessions
• Tool Integration
• AWS Bedrock

← 10. Cross-Platform Development | Next → 12. Testing with UniTest