📚 Nous AI Guides

Comprehensive guides for building production-ready AI agents with Nous. Grouped the same way as the HexDocs sidebar.

Core

Getting Started — install, first agent, tools, streaming, persistence.
Context & Dependencies — passing deps to tools/prompts and the run context.
Tool Development — building robust, validated, secure tools.
Structured Output — typed/validated results via Ecto schemas, schemaless types, or raw JSON schema.
Skills — reusable instruction/capability packages (module- and file-based) and the built-in catalog.
Hooks — lifecycle interceptors to block, modify, or audit agent actions.
Memory — persistent, searchable agent memory (keyword + vector, decay, scoping).
Knowledge Base — structured, linkable knowledge store and tools.
Evaluation — test suites, evaluators, metrics, and the optimizer.
Permissions & Guardrails — tool permission policies, approval gates, and session guardrails.
Observability & Telemetry — telemetry events, default handler, and Prometheus via PromEx.

Multi-Agent Teams — supervised agent groups, roles, shared state, comms, and rate limiting.
Decision Graph — track goals, decisions, and outcomes as a graph.
Deep Research — autonomous multi-step research with citations (Nous.Research.run/2).
Workflow Engine — DAG/graph-based orchestration of agents, tools, and control flow.

Providers Overview — all 13 providers + custom:, and how to switch with one string.
Custom Providers — the custom: OpenAI-compatible prefix.
HTTP Backends — Finch/Req vs Hackney, streaming backpressure.
Vertex AI Setup — Google Cloud Vertex AI configuration.
Fallback Chains — automatic provider/model failover.

Phoenix LiveView Integration — real-time streaming, multi-user coordination, production patterns.
Production Best Practices — architecture, security, scaling, monitoring.
Troubleshooting — common development and deployment issues.
Migration Guide — version upgrades and breaking changes.