- Refactor the `ResolveScript` function to improve the script path resolution strategy, allowing for better handling of assistant directories and module paths.
- Update asset metadata timestamps in `bindata.go` to reflect recent changes, ensuring accurate tracking of file modifications.
- Modify CSS component name formatting in `build.go` and TypeScript component selectors in `index.ts` for consistency in syntax.
- Improve locale handling in `locale.go` by implementing a fallback mechanism for locale file resolution based on language prefixes.
- Normalize locale values in `request.go` to lowercase for consistency across the application.
- Enhance component name sanitization in `utils.go` to replace additional characters, ensuring valid component naming conventions.
- Modify the test for the Claude executor to check for the IPC socket at the new path "/run/yao.sock" instead of the previous "/tmp/yao.sock". This aligns the test with the updated default configuration for the IPC socket path, ensuring consistency across the codebase.
- Modify the expected API URL in the `TestBuildProxyConfig` to include the `/v1` prefix for consistency with the updated URL building logic.
- Remove unnecessary `modified` flags in the `prepareAttachments` method to streamline attachment processing and ensure clarity in the codebase.
This change enhances the accuracy of tests and improves the maintainability of the attachment handling logic.
- Introduce a whitelist mapping for `package.yao` sandbox arguments to corresponding Claude CLI flags, allowing selective argument passing.
- Update `BuildCommandWithContinuation` to iterate over the whitelist and append valid arguments to the Claude CLI command.
This change enhances the flexibility of argument handling in the Claude CLI, ensuring only specified arguments are processed.
- Update `buildAPIURL` function in OpenAI provider to delegate URL construction to `connector.BuildAPIURL`, ensuring consistent URL formatting across the agent LLM and sandbox proxy paths.
- Modify backend URL construction in Claude's `BuildProxyConfig` to utilize the shared `connector.BuildAPIURL` helper, applying the necessary `/v1` prefix for compatibility.
This change enhances code maintainability and consistency in API URL handling across different components.
- Introduce functionality to resolve and manage user-uploaded files in the sandbox environment.
- Implement `prepareAttachments` method to convert attachment URLs to local file paths and handle duplicates.
- Update message processing to replace attachment content with text references, allowing Claude CLI to access files using Read and Bash tools.
- Enhance documentation to inform users about the new attachment handling capabilities.
This change improves the interaction with user-uploaded files, enabling better integration within the Claude CLI environment.
- Ensure the existence of $HOME/.Xauthority for PyAutoGUI/Xlib compatibility.
- Adjust PYTHONPATH to maintain access to user-installed Python packages.
- Explicitly set XAUTHORITY to the correct path for X11 authentication.
- Pass secrets as environment variables for Claude CLI to ensure proper access.
- Refactor VNC image support check to utilize a centralized keyword list for better maintainability.
- Fix ListDir to support BusyBox/Alpine ls by falling back from GNU
--time-style format, resolving CI test failures
- Update parseLS to handle both GNU (epoch) and BusyBox (date string) formats
- Fix claude-proxy streaming: always include usage in message_delta events
to prevent Claude CLI from falling back to non-streaming mode
- Fix claude-proxy non-streaming: ensure usage is always present in responses
- Add paragraph separators between text blocks in Claude executor stream parser
- Translate VNC proxy UI from Chinese to English
- Rename sandbox-claude-playwright to sandbox-claude-browser across
Dockerfiles, build scripts, and documentation
Co-authored-by: Cursor <cursoragent@cursor.com>
- Implement GetSandboxID method to return a mock sandbox ID for testing.
- Add GetVNCUrl method to return an empty string for VNC access in tests.
- Update SandboxExecutor interface to include new methods for sandbox identification and VNC URL retrieval.
- Enhance context creation to set sandbox ID and VNC URL properties in the sandbox instance.
Co-authored-by: Cursor <cursoragent@cursor.com>
- Update start command to check if the current directory is a Yao app root or a subdirectory, providing clearer error messages for users.
- Modify installation logic to handle empty directories more effectively, ensuring the init app is installed only when appropriate.
- Improve welcome message formatting and update links for documentation and community resources.
Co-authored-by: Cursor <cursoragent@cursor.com>
- Replace the existing NanoID generation in generateContextID with UUID for improved uniqueness.
- Update RequestID method to return the context ID directly instead of generating a new NanoID.
Co-authored-by: Cursor <cursoragent@cursor.com>
- Refactor BuildCommand to support session continuation using --continue flag.
- Introduce BuildCommandWithContinuation to handle both initial and continuation requests.
- Add BuildLastUserMessageJSONL for generating JSONL from the last user message in continuation scenarios.
- Enhance Executor to check for existing sessions and adjust command building accordingly.
- Update environment setup to ensure session data is stored correctly for persistence.
Co-authored-by: Cursor <cursoragent@cursor.com>
- Implement secure config file location (/tmp/.yao/proxy.json) instead of user-visible /workspace/
- Add generic options map support for backend-specific parameters (e.g., thinking for Volcengine GLM-4.7)
- Add secrets support for passing sensitive env vars (e.g., GITHUB_TOKEN) to sandbox container
- Remove excessive debug logs, keep critical ones with log.Printf("[Sandbox]...")
- Fix test assertions for system prompt passing via CLI args instead of env var
- Update i18n messages for sandbox loading states
Co-authored-by: Cursor <cursoragent@cursor.com>
- Send ChunkMessageStart before ChunkText in parseStream
- This initializes stream state (inGroup=true) required for Buffer.AddAssistantMessage
- Only send ChunkMessageEnd if message was started
Co-authored-by: Cursor <cursoragent@cursor.com>
- Fix Claude CLI output parsing: only extract content from final
assistant message (with stop_reason) to avoid duplicate content
- Remove trailing "..." from sandbox loading message
Co-authored-by: Cursor <cursoragent@cursor.com>
- Fix missing SystemPrompt field in sandbox executor options conversion
(was causing Claude CLI to be skipped even when prompts were configured)
- Rewrite parseStream to handle Claude CLI stream-json output format:
- system: initialization message
- assistant: message with content array (text, tool_use)
- result: final result with verification string
- Add comprehensive E2E tests via caller for sandbox integration:
- TestSandboxE2E_ClaudeCLIExecution: verify command execution
- TestSandboxE2E_FileCreation: verify file operations
- TestSandboxE2E_HookOnlyMode: verify Claude CLI skip logic
- TestSandboxE2E_StreamingResponse: verify streaming works
- Add real_e2e_test.go for direct Claude CLI execution testing
Co-authored-by: Cursor <cursoragent@cursor.com>
- Update TestClaudeCommandBuilding to expect new command format
- Rename TestClaudeCCRConfigBuilding to TestClaudeProxyConfigBuilding
- Update assertions for claude-proxy env vars instead of CCR
- Apply gofmt formatting to proxy/types.go
Co-authored-by: Cursor <cursoragent@cursor.com>
- Add --input-format stream-json, --output-format stream-json, --verbose flags
- Use heredoc to pass messages via stdin (no CLI length limit)
- Add BuildInputJSONL function for message conversion
- Add shouldSkipClaudeCLI logic to skip when no prompts/skills/mcp
- Update executor to conditionally start claude-proxy only when needed
- Add SystemPrompt field to Options for skip logic
- Add E2E tests for skip mode and command building
- Fix tests to reflect new command structure
Co-authored-by: Cursor <cursoragent@cursor.com>
- Add claude-proxy: lightweight Go proxy to translate between Anthropic
and OpenAI-compatible APIs with full streaming and tool-calling support
- Update Dockerfile to include claude-proxy binary (multi-arch)
- Add auto-start proxy via entrypoint when env vars are set
- Update executor.go to write proxy config and start proxy
- Simplify command.go to use direct Claude CLI with proxy
- Support both docker run -e and config file for proxy settings
Co-authored-by: Cursor <cursoragent@cursor.com>
- Adjusted assertions in the command test to verify the updated command structure, which now begins with "bash" and includes specific flags for executing CCR commands.
- Enhanced the test to ensure the user prompt is correctly included in the command, improving the accuracy of the test verification.
- Modified the command verification in the sandbox integration test to assert the new command structure, which now starts with "bash" and includes specific arguments for executing CCR commands.
- Enhanced assertions to check for the presence of required command components, improving the robustness of the test.
- Added functionality to build and manage MCP configuration for sandbox environments, allowing for dynamic tool execution.
- Enhanced the Assistant's Stream method to skip MCP tool calls in sandbox mode, with internal handling by Claude CLI.
- Introduced unit tests for MCP configuration building and skills directory resolution, ensuring robust integration.
- Updated sandbox manager to create IPC sessions and manage tool exposure dynamically, improving interaction with external agents.
- Enhanced documentation to reflect new features and integration points for MCP and skills within the sandbox.
- Updated the sandbox integration test to verify JSON fields using snake_case for CCR configuration.
- Added detailed documentation for the sandbox API, including properties, methods, and use cases for file operations and command execution.
- Enhanced context API documentation to include sandbox operations, improving clarity on available features when sandbox is configured.
- Added steps to pull necessary Docker images for sandbox testing in both CI workflows.
- Updated the AI test execution to utilize sandbox configurations, ensuring proper environment setup.
- Introduced sandbox initialization in the Assistant's Stream method, allowing for execution of coding agents like Claude and Cursor.
- Enhanced context management to support sandbox execution, improving flexibility in handling agent operations.
- Introduced an `OnCompleteCallback` type and updated the `Pool` struct to include an optional callback for execution completion.
- Implemented `SetOnComplete` method to allow setting the completion callback.
- Enhanced worker execution logic to notify the completion callback with the execution status (success, failure, or cancellation) after each execution, improving tracking and management of execution states.
- Updated execution control methods (`PauseExecution`, `ResumeExecution`, `StopExecution`) to include error handling and database status updates for paused, running, and cancelled states.
- Introduced `ExecuteWithControl` method in the Executor interface to manage execution with pause/resume capabilities, allowing for better control during execution.
- Enhanced the `Execute` methods across different executors (DryRun, Sandbox, Standard) to support pre-generated execution IDs and control mechanisms.
- Improved error messages in the execution handling API to provide clearer feedback on execution status and errors.
- Added support for a new execution status `ExecPaused` in the execution model, enhancing the tracking of execution states.
- Updated the expected default email channel in the `TestDefaultEmailChannel` test from "email" to "default" to align with recent configuration changes, ensuring accurate test validation.
- Updated the `buildEmailSubject` function to include the robot's display name in the email subject line, improving clarity and personalization.
- Modified the function signature to accept a `robot` parameter, allowing for dynamic subject prefixes based on the robot's identity.
- Ensured that the email subject reflects the robot's name when using template-based or summary-based subjects, enhancing user experience.
- Added new indirect dependencies: `github.com/JohannesKaufmann/dom` and `github.com/JohannesKaufmann/html-to-markdown/v2` for improved HTML processing.
- Updated `github.com/sergi/go-diff` to version 1.4.0 for enhanced diff functionality.
- Modified the `buildEmailBody` function to return both HTML and plain text versions of the email body, improving email formatting capabilities.
- Adjusted the default email channel name in configuration to "default" for better clarity.
- Implemented HMAC-SHA256 signature generation and verification for webhook requests to enhance security.
- Updated the Delivery Center to include the computed signature in the request headers when a secret is configured.
- Added utility functions for computing and verifying HMAC signatures, ensuring integrity of webhook payloads.
- Improved documentation for the new signature handling process to assist developers in implementing secure webhooks.
- Introduced manager ID and email fields in the Robot struct to facilitate mandatory delivery to managers.
- Updated delivery routing logic to ensure manager email is always included if manager ID is set, improving delivery reliability.
- Added JSON serialization handling in the Delivery Center to ensure compatibility with various data types.
- Enhanced context ID generation by replacing time-based IDs with NanoID for improved uniqueness and consistency.
- Introduced tests for validating MCP task output, ensuring only structure validation is performed without semantic checks.
- Implemented the `validateMCPOutput` function to validate MCP task outputs, checking for non-empty results and expected structure.
- Updated the `ExecuteWithRetry` method to incorporate MCP-specific validation logic, improving error handling and output validation for MCP tasks.
- Enhanced test coverage for various output scenarios, including nil, empty string, empty map, and empty array cases, to ensure robust validation behavior.
- Updated MCP task executor ID format to use "mcp_server.mcp_tool" for better clarity and consistency.
- Added required MCP-specific fields (`mcp_server` and `mcp_tool`) to the Task struct and validation logic.
- Enhanced documentation in DESIGN.md and TECHNICAL.md to reflect changes in MCP task structure and requirements.
- Improved error handling in ExecuteMCPTask to ensure proper validation of MCP task fields before execution.
- Updated database queries to include 'agents' and 'mcp_servers' fields in robot data retrieval.
- Enhanced the robot configuration structure to merge agents and MCP servers from the member table into the robot's resources.
- Improved the input formatter to display time markers with check/cross indicators for better context awareness.
- Added tests to validate the new functionality and ensure proper formatting of robot identity when identity is nil.
- Updated README files to include new commands for running tests and extracting results for review.
- Introduced the `yao agent extract` command to facilitate extraction of test results from JSONL files into Markdown or JSON formats.
- Enhanced the `FormatAvailableResources` function to support localization and detailed information for agents and MCP tools.
- Improved output formatting for better readability and usability in test result documentation.
- Added calls to CloseSafeWriter in multiple test cases to ensure all asynchronous writes are completed before verifying output.
- Improved test reliability by ensuring that the SafeWriter's buffer is properly flushed, preventing potential race conditions in output verification.
- Introduced mechanisms to handle agent-to-agent (A2A) calls, including automatic history skipping for forked calls and proper source tracking.
- Enhanced context management with the addition of ForkParentInfo to facilitate child stack creation without race conditions.
- Updated JSAPI methods to ensure correct handling of sub-agent calls, maintaining output isolation and preventing history pollution.
- Improved documentation to clarify the behavior of A2A calls and context management in concurrent scenarios.
- Implemented automatic disabling of SSE output for sub-agent calls to prevent message corruption and client disconnection.
- Introduced a `forceSkipOutput` method to ensure `skip.output` is set to true for all sub-agent requests.
- Updated the `Fork` method in the context to create independent Memory instances, preventing state sharing during concurrent executions.
- Enhanced documentation to clarify the behavior of SSE output and context management in batch operations.
- Introduced new methods `All`, `Any`, and `Race` to the MCP JavaScript API for concurrent tool calls across multiple MCP servers, enabling improved flexibility and performance.
- Updated existing methods to return parsed results directly, simplifying the response handling for developers.
- Enhanced documentation to include detailed descriptions and examples for the new cross-server operations, improving developer guidance and usability.
- Introduced `ctx.agent` and `ctx.llm` objects in the context API for improved agent-to-agent (A2A) and direct LLM calls.
- Added methods for single and parallel agent calls, including `Call`, `All`, `Any`, and `Race`, enabling flexible communication between agents.
- Enhanced documentation with detailed method summaries, parameters, and examples for better developer guidance on using the new APIs.
- Updated `logrus` dependency from v1.9.3 to v1.9.4 and `golang.org/x/sys` from v0.38.0 to v0.40.0 for improved functionality and security.
- Introduced a `Fork` method in the agent context to create child contexts for concurrent agent and LLM calls, preventing race conditions on shared state during batch operations.
- Enhanced the `Orchestrator` methods to utilize forked contexts, ensuring independent execution of agent calls without interference.
- Added initialization for the Agent JSAPI factory to support ctx.agent.* methods, improving agent interaction capabilities.
- Introduced a new agent object in the JSAPI context for calling other agents, enhancing modularity.
- Implemented an OnMessage callback in the context options to handle messages sent via ctx.Send(), allowing for more flexible message processing.
- Eliminated debug print statements for the "thinking" parameter and the full request body in the OpenAI provider, streamlining the request construction process.
- Maintained the functionality of including the "thinking" parameter for models that support reasoning/thinking mode, ensuring improved performance without unnecessary logging.
- Added support for a new "thinking" parameter in the request body for models that support reasoning/thinking mode, improving functionality.
- Updated the request body construction to include debugging output for the full request body, aiding in troubleshooting and development.
- Ensured that the model and other settings are retrieved from the connector, maintaining consistency in configuration handling.
- Introduced a new `Type` field in the `ActivityQuery` struct to allow filtering activities by type (e.g., execution.started, execution.completed, execution.failed, execution.cancelled).
- Updated the `ListActivities` method in `ExecutionStore` to handle the new type filter, mapping it to corresponding execution statuses.
- Enhanced unit tests in `execution_test.go` to validate filtering by activity type, including tests for valid and invalid type scenarios.
- Modified OpenAPI definitions and related types to support the new type filter in the activities endpoint, improving API usability and flexibility.
- Implemented `ListResults` and `CountResults` methods in `ExecutionStore` to retrieve and count completed executions with delivery content, supporting filtering by member ID, team ID, trigger type, and keyword.
- Developed `ListActivities` method to derive activities from execution status changes, allowing retrieval of recent activities for a team.
- Created corresponding unit tests in `execution_test.go` to validate the functionality of results and activities listing, ensuring accurate filtering and pagination.
- Updated OpenAPI routes and types to include new endpoints for listing results and activities, enhancing the API's capabilities for managing execution data.
- Enhanced the `extractGoalName` function to skip empty lines, markdown headers, and horizontal rules while extracting the first meaningful line as the goal name.
- Implemented a new `stripMarkdownFormatting` function to remove various markdown formatting elements, including bold, italic, inline code, and link syntax.
- Added comprehensive unit tests for both `extractGoalName` and `stripMarkdownFormatting` to ensure correct functionality across various markdown scenarios.
- Introduced a new `Description` field in the `Task` struct for a human-readable task description, improving UI clarity.
- Updated the `ParseTask` function to save the description from input data and convert it to a message if no explicit messages are provided.
- Enhanced the `Executor` to update UI fields with localized messages during task execution phases, ensuring better user feedback.
- Implemented a new method in the `ExecutionStore` to persist task status updates, allowing real-time UI updates.
- Added unit tests to validate the new task description handling and UI updates during execution phases.
- Refactored failure message construction to use concise phase names instead of full error messages, improving clarity in UI feedback.
- Updated localization support for failure messages, ensuring accurate translations for various phases in both English and Chinese.
- Added unit tests to validate the new failure message format and localization functionality.
