Guides

Each guide below covers one part of the Agent runtime in chai-ai. They build on each other roughly in this order:

1. Agents — defining tools, constructing an Agent, and how .generate() runs the tool-calling loop
2. Streaming — consuming the same loop as live events via .stream()
3. Sessions & Memory — persisting conversation history across calls
4. Structured Output — validated, typed final answers via outputSchema
5. Handoffs — transferring control to another agent mid-run
6. Guardrails — policy checks around input, tool calls, and output
7. Tracing — the runId/trace recorded on every run
8. Reliability — modelTimeoutMs / toolTimeoutMs