deathbyknowledge

This commit **enhances TypeScript type definitions** by **introducing a `Props` generic to the `ExportedHandler` interface** for `ExecutionContext`. This **new capability** allows for more flexible and type-safe handling of custom properties within the execution context across various handler types and their associated functions. The changes are applied to the core `src/workerd/api/global-scope.h` definitions and propagated to the `types/generated-snapshot/experimental/index.ts` and `types/generated-snapshot/latest/index.ts` type snapshot files, improving developer experience and code robustness.

This commit introduces a new example application for streaming chat using the OpenAI Agents SDK. It provides a new capability for developers to learn from and integrate, with no direct impact on existing production systems.

Fixes an environment type issue within the `hono-agents` middleware, ensuring its proper functionality. This is an isolated fix to a specific component.

This commit introduces a new capability to the `ai-chat` package, enabling AI chat agents to send and receive structured `data-*` stream parts. This significantly enhances the interactivity and data exchange capabilities for AI chat applications, offering new patterns for richer client-server communication.

This commit updates the scheduling module within the `agents` package, introducing API and schema changes due to an internal library migration. This is a breaking change for any services consuming the scheduling API.

This fix addresses an issue within the `agents` package, enabling callable methods to correctly return `this.state`. This improves the flexibility and correctness of state management for agents that utilize this pattern.

This commit introduces **WebSocket transport** as a new communication mechanism for the sandbox, enabling more efficient and persistent connections, particularly for **Cloudflare Workers/DO environments** by routing WebSocket upgrades through `stub.fetch()`. It involves a **major refactoring of the client-side transport layer** into a unified `ITransport` abstraction with dedicated `HttpTransport` and `WebSocketTransport` implementations, centralizing retry logic and standardizing the API. The **sandbox container** was updated with a `ws-adapter` to handle WebSocket upgrade requests and multiplex HTTP-like messages. Extensive **e2e test coverage** was added and configured in CI to validate all sandbox operations with both HTTP and WebSocket transports, significantly improving **robustness, error handling, and performance** through numerous bug fixes and CI/CD optimizations like parallelized Docker builds and caching.

This commit **restores and updates documentation** for **Anthropic agent patterns** within the `src/content/docs/agents/guides` section. It specifically modifies the `anthropic-agent-patterns.mdx` file to reflect a renewed focus on effective agent patterns. The **documentation update** includes changes to the document's title, external link, sidebar order, and overall description. This **content restoration** ensures users have access to relevant and up-to-date guidance for building robust Anthropic agents.

This commit **introduces a new comprehensive guide** on building **human-in-the-loop (HIL) AI agents** within the project's documentation. Located at `src/content/docs/agents/guides/human-in-the-loop.mdx`, this guide provides detailed steps, code examples, and best practices for implementing HIL patterns, including specific "anthropic patterns". This **enhances the project's educational resources** by offering users a dedicated resource to understand and apply HIL methodologies. The **new documentation feature** aims to improve user understanding and adoption of integrating human oversight into AI workflows, with minor content adjustments and link fixes also included.

This commit introduces a **fix** to prevent resource leaks and ensure proper termination of HTTP connections within the **sandbox** module. It modifies the `readStream`, `parseSSE`, and `parseSSEStream` functions across `interpreter-client.ts`, `file-stream.ts`, and `sse-parser.ts` respectively. The change ensures that **streams are explicitly cancelled or closed before their associated reader locks are released**, guaranteeing timely resource cleanup. This prevents hanging connections and improves the robustness and efficiency of stream processing, particularly for Server-Sent Events.

This commit fixes several scheduling and alarm-related bugs within the `agents` package, including immediate schedules and resource cleanup. It improves the robustness and correctness of scheduled operations within this module.

This commit **adds new documentation** to guide users on building and deploying interactive **ChatGPT Apps** using **Cloudflare Workers**. The new guide, located at `src/content/docs/agents/guides/chatgpt-app.mdx`, provides a comprehensive tutorial, including a practical example of a chess game application. This **enhances the developer experience** by offering clear instructions for integrating AI models with Cloudflare's edge platform, enabling users to leverage Workers for their AI-powered applications.

This commit **introduces a new guide** to the documentation, detailing the process of building and deploying an **AI-powered Slack Agent**. Specifically, it adds the `slack-agent.mdx` file under `src/content/docs/agents/guides/`, providing comprehensive instructions for integrating AI capabilities with Slack using **Cloudflare Workers**. This **documentation enhancement** empowers users to more easily develop and deploy intelligent agents that interact with their Slack workspaces, expanding the practical applications of Cloudflare's AI and serverless offerings.

This commit adds new documentation for the Agent class, improving clarity and understanding for developers. It is an isolated change to the documentation system with no impact on runtime behavior.

The `openai-sdk/chess-app` undergoes a significant architectural refactor, transitioning its core multiplayer platform (MCP) to a stateless worker model. This change impacts the application's operational efficiency, deployment configuration, and how it handles host information.

This commit refines the AI Playground user interface, improving the display of tools, fixing loading text, and preventing empty messages during tool calls. These changes enhance the user experience within the AI Playground.

This commit fixes state management and display issues within the AI Playground feature, specifically ensuring the MCP state resets correctly and debug logs are visible even when empty. This is an isolated change improving the user experience of the AI Playground.

This commit **adds a new tutorial** to the project's **documentation**, specifically within the **Sandbox** section. The new guide, located at `src/content/docs/sandbox/tutorials/claude-code.mdx`, provides step-by-step instructions on how to run **Claude Code** within the Sandbox environment. This **new content addition** enhances user understanding and facilitates the integration of the Claude Code model, improving the overall usability of the Sandbox for AI development.

This commit introduces a new, standalone `openai-sdk/chess-app` module, adding a significant new capability as a ChatGPT multiplayer chess application, isolated to its own application context.

This commit **introduces a new `claude-code` example** within the `examples` directory, showcasing a **Cloudflare Worker** that leverages a sandbox to execute **Claude Code**. This new capability demonstrates how to clone repositories, run tasks, and retrieve execution logs and diffs, providing a practical integration pattern for AI-driven code interpretation. Additionally, it includes a **maintenance fix** by updating development dependencies for `@types/node`, `typescript`, and `wrangler` in the existing `code-interpreter` example, ensuring type compatibility and stability. This expands the project's examples of AI integration and maintains existing ones.