## Summary
- Create CLAUDE.md as main development guide with architecture overview, code style guidelines, and setup instructions
- Add modular technical documentation in docs/claude/ directory covering:
- Database operations with query builder patterns
- Redis cache operations including pub/sub and distributed locks
- Stream operations for async producer-consumer patterns
- Generic types usage throughout the codebase
- HTTP request utilities and patterns
- Update README.md with documentation section for better discoverability
- Clean up unused cache functions in internal/db/cache.go
- Cross-reference related documentation for easy navigation
## Test plan
- [x] Verify all documentation links work correctly
- [x] Check code examples compile and follow established patterns
- [x] Ensure documentation reflects current codebase architecture
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
* feat(oauth): implement refresh credentials functionality
- Added RefreshCredentials endpoint to handle OAuth credential refresh requests.
- Introduced RequestOAuthRefreshCredentials structure for request validation.
- Updated access types and actions to include refresh credentials.
- Enhanced server routing and controller logic to support the new functionality.
- Updated OAuth entities to include expiration handling for refreshed credentials.
* feat(oauth): add metadata field to OAuthGetCredentialsResult
---------
Co-authored-by: Harry <xh001x@hotmail.com>
* feat(oauth): add RedirectURI field to OAuth request structures
* feat(oauth): update OAuthSchema validation
* feat: add Context field to request and session structures
* feat: add CredentialType field to Credentials and InvokeToolRequest structures
* fix: handle unhandled default case in basic_type.go
* feat: add support for build branches in build-push.yml
* refactor(plugin_manager): enhance HTTP client timeout handling in serverless runtime with DialContext
- Updated the HTTP client in the ServerlessPluginRuntime to use a context-aware DialContext for better timeout management.
- Removed the write timeout option from the HTTP request builder, streamlining the request handling process.
- Improved connection handling by setting write deadlines based on the PluginMaxExecutionTimeout.
This change enhances the reliability of network operations within the serverless runtime environment.
* refactor(plugin_manager): remove write deadline setting in serverless runtime connection initialization
- Eliminated the write deadline setting from the connection initialization in the ServerlessPluginRuntime.
- This change simplifies the connection handling process and aligns with the recent enhancements to timeout management.
This update contributes to a more streamlined and efficient network operation within the serverless environment.
* refactor(plugin_manager): adjust HTTP client timeout settings in serverless runtime
- Modified the HTTP client configuration in the ServerlessPluginRuntime to set the TLS handshake timeout based on PluginMaxExecutionTimeout.
- Retained the IdleConnTimeout setting to ensure consistent connection management.
This update improves the timeout handling for secure connections, enhancing overall network reliability in the serverless environment.
- Refactored the RemapAssets function to streamline the remapping of icon fields for both models and tools, including support for dark mode icons.
- Introduced new fields IconSmallDark and IconLargeDark in the ModelProviderDeclaration and added IconDark in ToolProviderIdentity to accommodate dark mode assets.
- Improved error handling during the remapping process for better clarity and maintainability.
* feat(plugin_manager): optimize local plugin startup with concurrent control
- Add semaphore-based concurrency control for plugin launches
- Implement parallel plugin startup using goroutines
- Optimize error handling to prevent goroutine blocking
- Add concurrency metrics logging
Note: handleNewLocalPlugins now accepts config parameter with default concurrency limit
* feat(plugin_manager): make local plugin launching concurrency configurable
* fix(plugin_manager): optimize comment and error handling
- Updated comments to clarify the concurrent plugin launching configuration.
- Added a nil check for the error channel during plugin startup to improve code robustness.
* refactor(plugin_manager): refactor plugin startup logic
- Remove the semaphore mechanism and switch to using routine.Submit for concurrency management
* fix(plugin_manager): Optimize plugin startup logs and concurrency control
- Added log output for maximum concurrency when starting local plugins
- Implemented a channel-based concurrency control mechanism to ensure limits are not exceeded
- Fixed closure variable capture issue to prevent incorrect plugin information
- Improved error handling to avoid deadlocks during startup
* fix(plugin_manager): simplify error channel handling and semaphore release logic
---------
Co-authored-by: jim02.he <jim02.he@vipshop.com>
* feat: add support for dark icon
- Introduced IconDark field in PluginDeclaration and related structures to support dark mode icons.
- Updated the installation process to handle dark icons.
- Enhanced asset validation to check for the presence of dark icons.
This change improves the visual consistency of plugins in dark mode environments.
* enhance(cli/icon): add plugin icon support with multiple categories
- Added support for light and dark icons for various plugin categories including agent, datasource, extension, model, tool, and trigger.
- Replaced the previous single icon implementation with a structured map for better organization and retrieval of icons based on category and theme.
- Removed the old Python icon file to streamline asset management.
This update improves the visual representation of plugins across different themes, enhancing user experience.
* change icons
* fix
* fix
* comments
- Implemented middleware to track active requests and active dispatch requests using atomic counters.
- Updated health check response to include counts of active requests and active dispatch requests.
- Integrated the new middleware into the HTTP server and plugin dispatch group for improved monitoring.
* feat: add InvokeLLMWithStructuredOutput functionality
- Introduced a new method InvokeLLMWithStructuredOutput to the BackwardsInvocation interface for handling structured output requests.
- Added corresponding request and response types to support structured output.
- Implemented the method in both RealBackwardsInvocation and MockedDifyInvocation for testing purposes.
- Updated permission handling and task execution for the new structured output invocation type.
This enhancement allows for more flexible and detailed responses from the LLM, improving the overall functionality of the invocation system.
* refactor: enhance LLMResultChunkWithStructuredOutput structure
- Updated the LLMResultChunkWithStructuredOutput type to include additional fields: Model, SystemFingerprint, and Delta.
- Added comments to clarify the reasoning behind the structure and the use of type embedding for JSON marshaling.
This change improves the clarity and functionality of the LLMResultChunkWithStructuredOutput type, ensuring proper JSON serialization.
* refactor: streamline LLMResultChunk construction in InvokeLLMWithStructuredOutput
- Simplified the construction of LLMResultChunk and LLMResultChunkWithStructuredOutput by removing unnecessary type embedding.
- Enhanced readability and maintainability of the code while preserving functionality.
This change contributes to cleaner code and improved clarity in the handling of structured output responses.
* feat(dynamic_select): implement dynamic parameter fetching functionality
- Added FetchDynamicParameterOptions function to handle dynamic parameter selection.
- Introduced new access type and action for dynamic select in access_types.
- Updated HTTP server routes to include the new endpoint for fetching dynamic parameters.
- Created necessary service and controller files for dynamic select operations.
* refactor(access_types): rename dynamic select access type to dynamic parameter
- Updated access type constants to reflect the change from PLUGIN_ACCESS_TYPE_DYNAMIC_SELECT to PLUGIN_ACCESS_TYPE_DYNAMIC_PARAMETER.
- Adjusted related references in the PluginDispatchers and FetchDynamicParameterOptions function to maintain consistency.
* feat: add decode plugin from identifier endpoint
- Introduced a new endpoint to decode a plugin from a unique identifier.
- Implemented the DecodePluginFromIdentifier function to handle decoding and verification of plugin signatures.
- Updated the HTTP server routes to include the new decode endpoint.
* refactor: update decode plugin endpoint path
- Moved the decode plugin from identifier endpoint to a new path for better organization.
- Updated the HTTP server routes accordingly to reflect the new endpoint structure.
- Updated .env.example to include DB_EXTRAS and DB_CHARSET variables.
- Refactored InitPluginDB functions for PostgreSQL and MySQL to accept a configuration struct, allowing for more flexible database connection settings.
- Adjusted connection pool settings to utilize new configuration options for charset and extras.
This change improves the configurability of database connections and prepares the codebase for future enhancements.
Directly fetch and invoke the listener for a given session_id instead of iterating over the entire listener map.
Signed-off-by: guanz42 <guanz42@163.com>
* bump cloud-kit version to fix#343
* change env name USE_AWS_S3 to S3_USE_AWS
* update s3_use_aws default value to true
* update readme about the upgrade notice
- Adjusted the header byte manipulation in chunking functions to correctly use the first four bytes for data length.
- Modified the HTTP request streaming function to include the length-prefixed option for improved data handling.
- Introduced new constants for HTTP option types to improve code readability and maintainability.
- Updated existing HTTP option functions to utilize the new constants.
- Implemented line-based and length-prefixed chunking methods for improved data processing in HTTP requests.
- Added comprehensive tests for chunking functionality to ensure reliability and correctness.
* refactor: update PluginManager to use configuration for various configurations
- Replaced hardcoded values in PluginManager methods with values from the configuration.
- Updated serverless plugin launch timeout and working paths to utilize the new configuration structure.
- Enhanced local plugin runtime initialization to pull settings from the configuration, improving maintainability and flexibility.
* refactor: clean up PluginManager by removing unused fields and updating platform check
- Removed commented-out fields from PluginManager to enhance code clarity.
- Updated platform check to utilize the configuration structure instead of a direct field reference, improving maintainability.
- Added `pluginMaxExecutionTimeout` to `PluginManager` for configurable execution limits.
- Updated `ServerlessPluginRuntime` to utilize the new timeout setting in HTTP requests.
- Refactored AWSPluginRuntime references to ServerlessPluginRuntime for consistency across the codebase.
- Updated the method in the interface to remove the parameter, simplifying its usage.
- Introduced a new function to provide a default verification structure.
- Added a file to store verification data, improving the plugin signing process.
- Enhanced tests in to validate the verification process, ensuring proper handling of success and failure scenarios.
- Refactored related code to accommodate the new verification structure and improve overall maintainability.
* fix: errChan failed to write response because of panic nil (#296)
* fix: join err and er into a single error using errors.Join, thanks @Yeuoly (#296)
---------
Co-authored-by: NeatGuyCoding <cto@sb>
* feat: Generate HTTP server routes from template
- Added a new file `http_server.gen.go` to automatically generate HTTP server routes based on defined dispatchers.
- Refactored existing route definitions in `http_server.go` to utilize the generated routes, improving maintainability.
- Introduced a code generation function in `generator.go` to create the HTTP server file, enhancing the plugin development workflow.
- Updated the template for HTTP server generation to streamline route creation for various controllers.
* fix: Update OAuth paths in PluginDispatchers for consistency
- Changed the path for authorization URL from `/oauth/authorization_url` to `/oauth/get_authorization_url`.
- Updated the path for credentials from `/oauth/credentials` to `/oauth/get_credentials` to align with naming conventions.
* feat: Add code generation for plugin controllers and services
- Introduced a code generation mechanism for plugin controllers and services, allowing for automatic generation based on defined dispatchers.
- Created new files for generated controllers, services, and templates to streamline the plugin invocation process.
- Removed outdated functions related to tool validation and runtime parameters, consolidating functionality into generated files.
- Updated dependencies in go.mod and go.sum to include necessary packages for the new code generation features.
* fix
- Introduced a new test case `TestLock` to validate the behavior of the Redis locking mechanism under concurrent access.
- Enhanced the `Lock` function to improve error handling and ensure proper locking behavior.
- Utilized `sync.WaitGroup` and atomic operations to measure wait times during lock acquisition, ensuring the lock behaves as expected under high concurrency.
* feat: Enhance plugin signing with authorized category verification
- Added support for an `authorized_category` flag in the signature command to validate the category before signing.
- Updated the `Sign` function to accept a verification parameter, allowing for category-based signing.
- Enhanced error handling for invalid categories during the signing process.
- Updated tests to cover new verification scenarios and ensure proper functionality with the authorized category.
* fix
* fix
* test
* test: Add unit test for plugin verification without verification field
- Introduced a new test case to verify the behavior of plugins that lack a verification field.
- Updated the signature_test.go file to include the test, ensuring proper functionality of the signing process.
- Removed the outdated verifier_test.go file and associated test data to streamline the codebase.
- Renamed parameters for consistency and clarity, changing `tenant_id`, `plugin_unique_identifier`, and `install_type` to `tenantId`, `pluginUniqueIdentifier`, and `installType` respectively across multiple functions.
- Updated corresponding database queries to reflect the new parameter names, enhancing code readability and maintainability.
* feat(redis): Add support for Redis Sentinel mode
Added support for Redis Sentinel mode to the Redis client, enabling automatic discovery and connection to the primary node through Sentinel. Updated relevant configuration files and initialization logic to support Sentinel mode configuration and connection.
* add lost RedisUser.
* feat: introduce 'run' command for local plugin execution
- Added a new command `run` to launch plugins locally, allowing communication through stdin/stdout.
- Removed the previous `test` command and its associated functionality to streamline the plugin testing process.
* feat: enhance 'run' command with TCP support
- Introduced a new `RunPluginPayload` structure to encapsulate plugin execution parameters.
- Added TCP communication mode to the `run` command, allowing multiple client connections.
- Updated command flags to configure run mode and logging options.
- Implemented client handling and server creation for both stdin/stdout and TCP modes.
* docs: improve comments in RunPlugin function for clarity
- Enhanced comments to provide clearer explanations of the plugin decoding process and the creation of client streams for both stdin/stdout and TCP modes.
- Updated comments to reflect the functionality and behavior of the plugin execution flow.
* refactor: update command structure and enhance plugin invocation handling
- Changed the command structure to add `runPluginCommand` under `pluginCommand` for better organization.
- Introduced `InvokePluginPayload` type to encapsulate plugin invocation details.
- Enhanced `RunPlugin` function to handle responses and errors more effectively, including logging to stdout.
- Updated client handling to support session management and improved error handling during plugin invocation.
- Renamed TCP server creation function for consistency.
* feat: enhance plugin response handling and logging capabilities
- Added a new flag `--response-format` to specify the output format (text or json) for plugin responses.
- Introduced a `logger` to manage logging output to stdout with timestamps and file information.
- Updated `logResponse` and `systemLog` functions to handle different response formats.
- Enhanced `handleClient` and `RunPlugin` functions to utilize the new response format feature.
- Implemented signal handling to clean up temporary directories on shutdown.
* feat: enhance plugin response structure and logging
- Added `InvokeID` to `InvokePluginPayload` and `GenericResponse` for better tracking of plugin invocations.
- Updated `logResponse` to include `InvokeID` in error responses for improved debugging.
- Enhanced client handling in `handleClient` to log plugin readiness and received requests.
- Refactored client stream creation for better readability and consistency.
* feat: add plugin invoke end response type and logging
- Introduced `GENERIC_RESPONSE_TYPE_PLUGIN_INVOKE_END` to enhance response tracking for plugin invocations.
- Updated `handleClient` to log the end of plugin invocation, improving visibility into the plugin lifecycle.
* chore: remove fullfeature tags
- Introduced a new test file to implement a fake OpenAI server that streams responses for benchmarking purposes.
- Updated the existing benchmark test to utilize the new fake server, enhancing the testing environment for local plugin runtime invocations.
- Refactored the package name in the benchmark test file for clarity and consistency.
* feat: add plugin reinstallation functionality and admin API support
- Implemented ReinstallToAWSFromPkg method to allow reinstallation of plugins on AWS Lambda, updating function URL and name.
- Added clearServerlessRuntimeCache method to manage serverless runtime cache.
- Enhanced LaunchPlugin to support an ignoreIdempotent flag for forced reinstallation.
- Introduced admin API endpoints for plugin reinstallation, secured with an API key validation middleware.
- Updated configuration to include AdminApiEnabled and AdminApiKey settings.
* refactor: update plugin reinstallation endpoint and improve unauthorized response
- Changed the plugin reinstallation endpoint from "/plugins/reinstall" to "/plugin/serverless/reinstall" for better clarity.
- Modified the unauthorized response in the AdminAPIKey middleware to return a more descriptive JSON message.
- Implemented validation in NewInstalledBucket to log warnings for non-alphanumeric starting characters and empty paths for installedPath.
- Updated parameter naming for consistency across methods in InstalledBucket, enhancing code clarity.
- Adjusted the buffer size settings in the stdioHolder to utilize configurable values for stdout buffer size and maximum buffer size, enhancing flexibility in plugin output handling.