Gabriel Massadas

This commit introduces **new capabilities** by adding comprehensive support for **AI Search namespace and instance binding types** across the Cloudflare Workers ecosystem. It enables developers to configure and utilize `ai_search_namespaces` and `ai_search` bindings in their Workers, integrating with Cloudflare's AI Search service. The changes span **Wrangler** for deployment, configuration, and local development, **Miniflare** for local emulation, and **workers-utils** for core type definitions and validation. This feature allows for seamless interaction with AI Search resources, enhancing the functionality of Workers both in production and during local development.

This commit **updates and adds documentation** for the **AI Search** product, specifically focusing on **custom metadata filtering**. It provides comprehensive guides on configuring custom metadata schemas, including examples for R2 and website data sources, and details built-in attributes and supported data types. The changes also include updates to the limits and pricing documentation to reflect new constraints for custom metadata fields. This **documentation update** ensures users have the necessary information to leverage **custom metadata filtering** for more precise search capabilities and data organization within **AI Search**.

This commit **activates RPC routing** for the `autorag()`, `gateway()`, and `toMarkdown()` functions within the **Cloudflare AI API**, consolidating their experimental access under the existing `experimental` compatibility flag. This **new feature** streamlines the management of these experimental RPC paths, with `gateway()` additionally supporting an `options.beta` flag for activation. The core logic resides in `src/cloudflare/internal/ai-api.ts`, and the change is accompanied by **new testing infrastructure** for `autorag` RPC, including build configurations, a dedicated test, and a mock.

This commit introduces a **new feature** by adding an optional `beta` flag to the `gateway` method within the **AI API client** (`src/cloudflare/internal/ai-api.ts`). This flag enables **conditional routing of AI gateway methods via RPC**, allowing for alternative communication pathways. The underlying `Fetcher` interface was also updated to support this new routing mechanism. This change provides flexibility for testing or deploying AI gateway functionalities through a different backend infrastructure, potentially impacting performance or reliability for specific users.

This is an isolated change to the frontend's AI Search component, refactoring its filter type definitions for improved consistency and maintainability by re-using existing vectorize types. It has no direct user-facing impact.

This commit **fixes a type definition error** within the **AI Search functionality** by correcting the TypeScript type signature for `aiSearch`. Previously, `aiSearch` was incorrectly defined as a property, but this change updates it to a callable method in both the `experimental` and `latest` API snapshots. This **bug fix** ensures that developers using TypeScript can correctly invoke the **AI Search API**, preventing compilation errors and improving the overall developer experience.

This commit **introduces new RPC types** specifically designed to support the **AI Search feature**. It defines the necessary data structures and interfaces for remote procedure calls, enabling robust and type-safe communication between services that will interact with the AI-powered search functionality. This work represents a **new capability enablement**, laying the foundational infrastructure for the development and integration of advanced AI search capabilities within the system. The addition of these types ensures proper data serialization and deserialization, facilitating future implementation and expansion of AI search operations.

Fixes AI Search RPC binding issues within `miniflare` and `wrangler`, ensuring correct integration and functionality for developers utilizing AI features.

This commit introduces a new AI Search RPC within the internal Cloudflare AI API, enabling new AI-powered search capabilities for internal services.

This commit enhances `wrangler`'s user authentication system by adding new OAuth scopes specifically for AI Search. This introduces a new capability, allowing `wrangler` to request necessary permissions for future AI Search-related deployments or interactions.

This commit introduces significant **new capabilities** and **documentation enhancements** for **AI Search**, focusing on improved data ingestion and API interaction. It enables **path filtering** for both **website and R2 data sources**, allowing users to precisely control which content is indexed, and simplifies **AI Search instance creation** through a new REST API. Comprehensive documentation for the new **service API token** is also added, alongside updates to changelogs and release notes reflecting these improvements and the introduction of a new indexing engine and hybrid search. These changes enhance user control over data ingestion and streamline programmatic interaction with the **AI Search** platform.

This commit deprecates the optional `gateway option id` within internal universal requests, affecting the `aig-api` and its generated type definitions, and requiring internal consumers to adapt to the evolving API contract.

This commit introduces a new `system_prompt` field to Cloudflare's internal RAG AI search API, enhancing its capability to guide AI search behavior. This is an isolated new feature for the RAG AI service.

This commit fixes a bug in the AI Gateway, ensuring that extra headers are correctly applied even when no other options are provided, restoring full functionality for users needing to send custom headers.

This commit delivers a **bug fix** by making the `autoragId` parameter **optional** within the `autorag` method signature. It specifically updates the type definitions in `types/generated-snapshot/experimental/index.ts` and `types/generated-snapshot/latest/index.ts` to reflect this change. This resolves an issue where the `autorag` class's `id` property was incorrectly required during initialization, improving the **flexibility and usability** of the **`autorag` API**. Developers can now invoke the `autorag` method without explicitly providing an `autoragId`, streamlining its integration and preventing unnecessary errors.

This commit performs a **refactoring** and **API schema update** for the **AutoRAG API client**. It primarily **removes the `source_params` field** from the `AutoRagListResponse` type definition, affecting both the `experimental` and `latest` API snapshots. Additionally, the `AutoRAG` constructor in `src/cloudflare/internal/autorag-api.ts` has been updated to accommodate an optional `autoragId`. This change streamlines the data structure returned by the AutoRAG list endpoint, ensuring that consumers of this API no longer expect or receive the `source_params` field.

This commit **enhances the AI Gateway documentation** by adding a new section to the `universal.mdx` page. It introduces a **Workers Binding example**, illustrating how to integrate the **AI Gateway** with Cloudflare Workers. This **documentation improvement** helps users understand and implement Workers Bindings for AI Gateway, making the feature more accessible and easier to adopt for their projects.

This commit introduces a **new capability** to the **AutoRAG binding**, enabling users to list available RAG configurations. A new `list` method has been added to the `AutoRAG` class within `src/cloudflare/internal/autorag-api.ts` to facilitate this functionality. To support the listing feature and enhance flexibility, the `autoragId` parameter has been made optional in the `autorag` method in `src/cloudflare/internal/ai-api.ts`, as well as in the `AutoRAG` constructor, `search`, and `aiSearch` methods. Additionally, a new `AutoRAGNameNotSetError` is introduced for improved error handling.

This commit **introduces filtering capabilities** for **AutoRAG search requests**, enhancing the precision and relevance of information retrieval. It is a **new feature** that defines new data structures, `ComparisonFilter` and `CompoundFilter`, and integrates a `filters` field into the `AutoRagSearchRequest` within the `src/cloudflare/internal/autorag-api.ts` file. This allows users to apply specific criteria to their AutoRAG queries, enabling more targeted and effective searches within the **AutoRAG API**.

This commit introduces a **new documentation page** for **AutoRAG usage recipes** at `src/content/docs/autorag/usage/recipes.mdx`, offering practical examples for implementing custom models and simple search engines. As a **documentation enhancement**, it also updates the `src/content/docs/autorag/configuration/indexing.mdx` file with an improved size limit link. Minor style adjustments were also applied to `src/content/products/autorag.yaml`. This work significantly improves the **AutoRAG documentation** by providing clearer guidance and examples for users.