Engine-only profile registries for resolving `InferenceSettings` in Geppetto.

Engine-first event publishing with EventSink, Watermill routing, and context-carried sinks for streaming AI.

Task-based index for Geppetto docs and example programs.

App-facing `pkg/inference/runner` package for building simple and event-driven inference programs from resolved runtime input.

Step-by-step guide to implement a custom event handler, subscribe it to the router, and parse incoming events.

Step-by-step guide to define a tool, register it, attach the registry to context, and configure tool calling on a Turn.

Applications now bootstrap StepSettings from defaults/config/env/flags only; profiles no longer patch them.

A practical guide to writing, composing, and using middlewares with Turn-based engines.

Legacy profile-map migration has been removed after the StepSettingsPatch hard cut.

Step-by-step guide to migrate from the legacy inference lifecycle APIs to geppetto/pkg/inference/session.

Runbook for operating, inspecting, backing up, and recovering SQLite-backed profile registry storage.

Step-by-step guide to extract structured data from streaming LLM output using filtering sinks and custom extractors.

Managing multi-turn interactions with session.Session — turn history, inference lifecycle, and async execution.

A complete guide to defining, attaching, and executing tools with Turns. Tool registries are carried via `context.Context` (Turn state stays serializable).

Understanding the Run/Turn/Block data model and how engines and middleware use it

A guide to the embeddings package in Geppetto, covering embedding generation, caching, and integrating with applications.

A comprehensive guide to the Turn-based inference engine architecture, covering engines, streaming, and provider implementations.

Resolve provider credentials through Geppetto/Pinocchio profiles and pass final settings to engines.