ai-query is a Python SDK for building AI features at three levels:
- direct model calls
- stateful agents
- live controllable runtimes
The Ladder
First Example
Learn In Order
Quickstart
Build from one model call to a stateful agent in a few minutes.
Concepts
Understand the three layers: core generation, agents, and live turns.
Navigating ai-query
Choose the right primitive for your app.
Generating Text
Learn the raw model call before adding state or servers.
Core Building Blocks
Streaming
Stream model output directly or through an agent.
Tools
Expose typed Python functions to models.
Tool Loops
Understand multi-step tool execution and stop conditions.
Providers
Configure OpenAI, Anthropic, Google, OpenRouter, and more.
Stateful Agents
Stateful Agents
Add identity, storage, message history, and state.
Conversations
Work with message history and multi-turn context.
Live Turns
Build CLIs, UIs, SSE streams, and steerable runtimes.
Agent Server
Route many users to persistent agents.
Reference And Recipes
API Reference
Look up classes, functions, types, transports, and providers.
Cookbook
Copy working patterns for RAG, tools, MCP, deployment, and workflows.