#Introduction

#Llama Stack

Llama Stack is a framework for building and running AI agents with tools. It provides a server-based architecture that enables developers to create agents that can interact with users, access external tools, and perform complex reasoning tasks.

Main components and concepts include:

• Llama Stack Server: Central service that hosts models, agents, and tool runtime. It can be deployed on Kubernetes via the Llama Stack Operator (see Install Llama Stack).
• Client SDK (llama-stack-client): Python client for connecting to the server, creating agents, defining tools with the @client_tool decorator, and managing sessions.
• Agents: Configurable AI agents that use LLM models and can call tools (e.g., weather API, custom APIs) to answer user queries.
• Tools: Functions exposed to the agent (e.g., weather query). Defined with @client_tool and passed to the agent at creation time.
• Configuration: YAML stack configuration defines providers (inference, agents, safety, vector_io, files), persistence backends, and model registration (e.g., DeepSeek via OpenAI-compatible API).

Llama Stack supports multiple API providers, storage and persistence backends, and distribution options (e.g., starter, postgres-demo, meta-reference-gpu), making it suitable for quick experiments and production deployments.

#Documentation

Llama Stack provides official documentation and resources for in-depth usage:

#Official Documentation

• Main Documentation: https://llamastack.github.io/docs
  • Usage, API providers, and core concepts
• Core Concepts: https://llamastack.github.io/docs/concepts
  • Architecture, API stability, and resource management