Higress v2.1.5: 41 engine updates + 8 console updates

Higress Community

Jul 23, 2025

Share on X

Higress Engine Update

📋 Overview of This Release

This release includes 41 updates, covering various aspects such as feature enhancements, bug fixes, performance optimizations, and more.

Distribution of Updates

New Features: 19 items
Bug Fixes: 14 items
Refactoring Optimizations: 2 items
Documentation Updates: 6 items

⭐ Key Focus

This release includes 2 significant updates that are recommended for close attention:

feat: add DB MCP Server execute, list tables, describe table tools (#2506): By adding these tools, users can manage and operate databases more conveniently, enhancing the system's flexibility and usability, making database operations more intuitive and efficient.
feat: advanced load balance policies for LLM service through wasm plugin (#2531): By introducing advanced load balancing strategies, the performance and resource utilization of the LLM service have been improved, allowing users to choose the most suitable strategy to optimize their services based on demand.

For detailed information, please refer to the important features section below.

🌟 Detailed Description of Important Features

Here are detailed explanations of the important features and improvements in this release:

1. feat: add DB MCP Server execute, list tables, describe table tools

Related PR: #2506 | Contributor: hongzhouzi

Background

In many application development scenarios, developers need to frequently interact with databases, such as executing SQL statements, viewing table structures, etc. The existing MCP server supports basic database query functions but lacks more advanced operational tools. This update adds execute (execute SQL), list tables (list tables), and describe table (describe table) tools, aiming to meet users' higher demands for database management. The target user group includes but is not limited to database administrators, backend developers, and application developers who frequently interact with databases.

Feature Details

In terms of specific implementation, new constant types for the database have been introduced by modifying the db.go file, and new tools have been registered in server.go. The new tools enable executing any SQL statements, listing all table names, and obtaining detailed information about specific tables. The core technical focus is on using the GORM framework to handle different types of database connections, while providing customized SQL query logic for each type of database. Additionally, code changes also involve optimizing the error handling mechanism, such as unifying the error handling function handleSQLError, improving code maintainability. These improvements not only enrich the function set of the MCP server but also enhance its applicability in various database environments.

Usage

Enabling these new features is very simple. Just ensure your MCP server configuration includes the correct database DSN and type. For the execute tool, users can execute INSERT, UPDATE, or DELETE operations by sending a request that contains the sql parameter; the list tables tool does not require any additional parameters, simply calling it returns all table names in the current database; while the describe table tool requires a table parameter to specify the table name whose structure to view. Typical usage scenarios include but are not limited to: regularly checking the consistency of the database table structure, automated script generation, and validation before and after data migration. It is important to note that when using the execute tool, caution is necessary to avoid executing commands that may compromise data integrity.

Feature Value

This feature greatly expands the MCP server's application scope in database management, allowing users to more efficiently complete daily tasks. It simplifies the complex manual operation process, reduces the probability of errors, and also provides a solid foundation for building automated operation and maintenance processes. Especially for projects that need to work across multiple database platforms, this unified and flexible interface design is undoubtedly a boon. In addition, by improving error handling logic and adding security measures (such as preventing SQL injection), this PR further ensures the system's stability and security.

2. feat: advanced load balance policies for LLM service through wasm plugin

Related PR: #2531 | Contributor: rinfx

Background

With the widespread application of large-scale language models (LLM), the demand for high performance and high availability is growing. Traditional load balancing strategies may not meet this demand, especially when handling a large number of concurrent requests. The new load balancing strategy aims to address these issues, providing a more intelligent way to allocate requests. The target user group includes enterprises and developers requiring high-performance and high-availability LLM services.

Feature Details

This PR implements three new load balancing strategies: 1. Minimum load policy, implemented based on WASM, suitable for gateway-api-inference-extension; 2. Global minimum request count policy based on Redis, tracking and managing the request count for each host through Redis to ensure requests are assigned to the current least loaded host; 3. Prompt prefix matching strategy, selecting backend nodes based on the prompt prefix, if no match is found, the global minimum request count policy is used. These strategies are implemented via WASM plugins, providing high scalability and flexibility.

Usage

To enable these load balancing strategies, you need to specify the corresponding policy type and configuration parameters in the Higress gateway configuration. For example, to enable the global minimum request count policy based on Redis, you need to set lb_policy to global_least_request, and provide the FQDN, port, username, and password for the Redis service. For the prompt prefix matching strategy, you also need to set lb_policy to prefix_cache, and make the corresponding configuration. The best practice is to choose the appropriate strategy based on actual application scenarios and regularly monitor and adjust configurations to optimize performance.

Feature Value

These new load balancing strategies bring significant performance improvements to LLM services. The minimum load strategy ensures requests are assigned to the current least loaded host, thereby improving response speed and resource utilization. The global minimum request count policy based on Redis further optimizes resource allocation through real-time tracking of each host's request count. The prompt prefix matching strategy improves processing efficiency by caching and reusing KV Cache. These features not only enhance the system's performance and stability but also improve user experience, especially under high concurrency scenarios.

📝 Complete Change Log

🚀 New Features (Features)

Related PR: #2533 Contributor: johnlanni Change Log: Added support for subPath field, allowing users to configure rules for removing request path prefixes, as well as updated the Chinese and English documentation to include usage instructions for the new feature. Feature Value: By introducing the subPath configuration option, it enhances the flexibility and customizability of the AI agent plugin, allowing developers to have more granular control over request path handling logic, improving user experience.
Related PR: #2514 Contributor: daixijun Change Log: This PR commented out the default configuration of tracing.skywalking in values.yaml, addressing the issue of automatically adding skywalking configurations when other tracing types are chosen by users. Feature Value: By removing unnecessary skywalking configurations, it avoids conflicts with user-defined tracing settings by default, enhancing system flexibility and user experience.
Related PR: #2509 Contributor: daixijun Change Log: This PR implements the processing of OpenAI responses interface Body and adds support for the Volcano Ark large model responses interface by expanding the logic in provider/doubao.go. Feature Value: The new functionality enables the system to support more types of AI response processing, significantly improving compatibility and flexibility for users using the Volcano Ark large model.
Related PR: #2488 Contributor: rinfx Change Log: Added trace_span_key and as_separate_log_field configuration items, allowing the log records and span property record key to differ, and allowing log content to exist as an independent field. Feature Value: By providing more flexible logging and tracing data recording methods, it enhances the system monitoring capability, helping developers better understand and optimize application performance.
Related PR: #2485 Contributor: johnlanni Change Log: This PR introduces the errorResponseTemplate functionality, allowing the MCP server plugin to customize response content when the backend HTTP status code is greater than 300. Feature Value: This feature allows users to customize error response templates based on actual situations, enhancing system flexibility and user experience, especially providing friendlier feedback when handling exceptional situations.
Related PR: #2460 Contributor: erasernoob Change Log: This PR modifies the message endpoint sending logic of the SSE server in the mcp-session plugin, enabling it to pass query parameters to the REST API server, and processes the sessionID with URL encoding. Feature Value: By supporting passing query parameters from the SSE server to the REST API server, it enhances the system's flexibility and functional integration, making it easier for users to customize service requests.
Related PR: #2450 Contributor: kenneth-bro Change Log: Added a sector market MCP Server, integrating the latest real-time market data and component stock information for industry and concept sectors. Feature Value: Provides users with detailed market data analysis tools to help investors track the performance of industry and concept sectors in real-time, leading to more informed investment decisions.
Related PR: #2440 Contributor: johnlanni Change Log: This PR fixes two issues in Istio and Envoy, and adds a new WASM API to support injecting encoding filter chains during the encodeHeader phase. Feature Value: By solving issues related to consistent hashing and providing new APIs, this update improves the system's stability and flexibility, allowing users to have more granular control over the request handling process.
Related PR: #2431 Contributor: mirror58229 Change Log: This PR adds default route support for wanx images and video synthesis, and updates related README files to reflect these changes. Feature Value: By introducing default route support, users can handle wanx image and video synthesis requests more flexibly, enhancing system usability and user experience.
Related PR: #2424 Contributor: wydream Change Log: This PR adds support for OpenAI Fine-Tuning API in the ai-proxy plugin, including path routing, capability configuration, and related constant definitions. Feature Value: By introducing support for Fine-Tuning API, users can now utilize this service for more advanced model fine-tuning tasks, enhancing system flexibility and functionality.
Related PR: #2409 Contributor: johnlanni Change Log: Added a WASM-Go plugin named mcp-router that supports dynamic routing of MCP tool requests, including the creation of Dockerfile, Makefile, and related documentation. Feature Value: This plugin allows aggregation of different tools from multiple backend MCP servers through a single gateway endpoint, streamlining multi-service integration and management, enhancing the system's flexibility and scalability.
Related PR: #2404 Contributor: 007gzs Change Log: This PR added support for reasoning_content for the AI data mask function and supports returning multiple index groups in the request, enhancing the flexibility and diversity of AI responses. Feature Value: By enhancing support for reasoning_content and allowing multiple index groups to be returned, users can more flexibly handle AI response data, improving application adaptability and user experience in complex scenarios.
Related PR: #2391 Contributor: daixijun Change Log: Adjusted the streaming response structure of the AI agent to ensure that null is output for the usage, logprobs, and finish_reason fields when they are empty, maintaining consistency with the OpenAI interface. Feature Value: By maintaining consistency with the OpenAI interface, it improves the system's compatibility and user experience, allowing developers to integrate and use the API more conveniently.
Related PR: #2389 Contributor: NorthernBob Change Log: This PR implements one-click deployment support for plugin servers on Kubernetes and configures the default download URL for plugins. Changes include adding and modifying multiple Helm template files to support plugin servers. Feature Value: By supporting one-click deployment on Kubernetes and preset plugin download URLs, it simplifies the user's process of deploying and using plugins in the K8s environment, improving usability and efficiency.
Related PR: #2378 Contributor: mirror58229 Change Log: This PR adds support paths for WANXIANG image/video generation in ai-proxy and introduces a configuration item in ai-statistics to avoid errors related to OpenAI. Feature Value: Provides users with new image and video generation capabilities, while ensuring system stability and compatibility through new configuration items, enhancing user experience.
Related PR: #2343 Contributor: hourmoneys Change Log: This PR introduces an AI-based bidding information tool MCP service, including detailed English and Chinese README files and configuration descriptions. Feature Value: The new feature allows users to query the bidding information list through keywords, enhancing the ability of enterprises to obtain projects and clients, providing more comprehensive and accurate information support.
Related PR: #1925 Contributor: kai2321 Change Log: This PR implements the AI-image-reader plugin, which parses image content by interfacing with OCR services (such as Alibaba Cloud Lingji). It adds related Go code and bilingual documentation. Feature Value: This function allows users to automatically read and process text information from images using AI technology, enhancing the system's intelligence level and user experience.

🐛 Bug Fixes (Bug Fixes)

Related PR: #2524 Contributor: daixijun Change Log: This PR fixes the issue of stream_options parameters being misused on non-openai/v1/chatcompletions endpoints by limiting this parameter to only be effective on designated endpoints to avoid errors. Feature Value: Ensures the correctness of API calls, prevents errors caused by parameter misuse, and enhances system stability and user experience.
Related PR: #2516 Contributor: HecarimV Change Log: This PR enhances the AI Proxy component by adding system message processing capability to Bedrock API requests, specifically implementing the inclusion of the System field in the request body structure and updating the request construction logic to conditionally include system messages. Feature Value: Enhances the AI agent's support for Bedrock services, allowing users to attach system-level instructions or information when sending requests, which helps to more precisely control the style and direction of the generated content, improving user experience and application flexibility.
Related PR: #2497 Contributor: johnlanni Change Log: This PR fixes the incorrect decoding behavior when the configured URL path contains URL-encoded parts by modifying the lib-side code. Feature Value: This fix ensures the correct decoding when processing request paths containing URL-encoded parts, improving system stability and user experience.
Related PR: #2480 Contributor: HecarimV Change Log: This PR fixes the issue of AWS Bedrock supporting additional request fields, ensuring that the AdditionalModelRequestFields field is correctly initialized, avoiding potential null pointer exceptions. Feature Value: By increasing support for additional model request fields, users can configure AWS Bedrock services more flexibly, enhancing API call customization capabilities and stability.
Related PR: #2475 Contributor: daixijun Change Log: Fixes the 404 problem that occurs when the openaiCustomUrl is configured for a single interface with a path prefix that is not /v1 due to incorrect customPath passing. Adjustments to request handling logic ensure compatibility. Feature Value: This fix resolves the 404 error users encounter under specific conditions, enhancing stability and user experience when using a customized OpenAI service path.
Related PR: #2469 Contributor: luoxiner Change Log: Fixes the problem of excessive log generation during the discovery process of the MCP server when Nacos is unavailable by rectifying erroneous log record calls to reduce unnecessary log output. Feature Value: Reduces the amount of logs generated by the system when Nacos service is unreachable, avoiding storage pressure and performance issues caused by rapidly growing log files, enhancing system stability and user experience.
Related PR: #2445 Contributor: johnlanni Change Log: Fixes the issue where the MCP server does not return a body when responding with a status; now uses SSE responses; also refactored makeHttpResponse. Feature Value: Addresses potential errors caused by a lack of a response body, improving system stability and user experience, ensuring correct communication between the backend and frontend.
Related PR: #2443 Contributor: Colstuwjx Change Log: This PR fixes a problem by adding missing annotations in the controller service account, allowing users to set annotations for the controller service account. Feature Value: This change allows users to configure service accounts more flexibly, such as binding AWS IAM roles to service accounts through annotations to authenticate against AWS resources.
Related PR: #2441 Contributor: wydream Change Log: This PR standardizes the naming conventions for API name constants and corrects mapping errors in the getApiName function, ensuring that API requests can match correctly. Feature Value: By correcting inconsistencies in API name spellings and formats, it enhances system stability and reliability, avoiding functional failures or 404 errors caused by path errors.
Related PR: #2423 Contributor: johnlanni Change Log: This PR resolves a potential crash issue of the controller when configuring the MCP server for SSE forwarding by modifying the relevant logic in ingress_config.go to prevent exceptional situations. Feature Value: Fixes the potential crash issue of the controller, enhancing system stability and reliability, ensuring users won't encounter service interruptions when using SSE forwarding features.
Related PR: #2408 Contributor: daixijun Change Log: Adjusts the finishReason returned by the Gemini API to lowercase and fixes missing finishReason content in streaming responses, ensuring consistency and completeness with the OpenAI API. Feature Value: This fix enhances the API's compatibility and stability, ensuring that users can receive consistent and complete response results when using the Gemini provider, improving user experience.
Related PR: #2405 Contributor: Erica177 Change Log: Corrects the spelling error of McpStreambleProtocol, ensuring that protocol support logic, type mapping, and routing rewrite rules are correct. Feature Value: Fixes issues with protocol identification and mapping due to constant name spelling errors, improving system stability and reliability.
Related PR: #2402 Contributor: HecarimV Change Log: Fixes the Bedrock Sigv4 signature mismatch issue in the AI agent, improving the modelId decoding logic to avoid potential data contamination risks. Feature Value: This fix enhances system stability, preventing failures in service calls due to incorrect model IDs, improving user experience and system reliability.
Related PR: #2398 Contributor: Erica177 Change Log: Corrects the spelling error in the constants of McpStreambleProtocol, changing it from ‘mcp-streamble’ to ‘mcp-streamable’, and adjusts related references to ensure consistency and correctness of protocol names. Feature Value: Fixes potential issues with protocol matching failures or configuration parsing problems due to spelling errors, enhancing system stability and reliability, avoiding service anomalies caused by such simple faults.

♻️ Refactoring (Refactoring)

Related PR: #2458 Contributor: johnlanni Change Log: This PR updates the dependency of the MCP server to the latest version of the wasm-go repository, adjusting the dependency paths in go.mod to ensure the project uses the latest codebase. Feature Value: By relying on the latest wasm-go repository, users can ensure the project utilizes the latest features and performance optimizations, enhancing system stability and compatibility.
Related PR: #2403 Contributor: johnlanni Change Log: This PR standardizes newline character markers in the MCP session filters, achieving consistency by modifying two code lines in sse.go. Feature Value: Standardizing newline character markers reduces confusion due to formatting inconsistencies, improves code readability and maintainability, making it easier for developers to understand and use related features.

📚 Documentation Updates (Documentation)

Related PR: #2536 Contributor: johnlanni Change Log: This PR mainly updates version numbers and version information in related configuration files to prepare for the release of version 2.1.5. Feature Value: By updating the version number to reflect the latest software status, users can clearly understand the current software version and its stability.
Related PR: #2503 Contributor: CH3CHO Change Log: Corrected the spelling error of the configuration item name in the ai-proxy plugin README, changing vertexGeminiSafetySetting to geminiSafetySetting. Feature Value: Ensures the accuracy of documentation, avoiding users being unable to set configurations correctly due to incorrect item names, improving user experience and document readability.
Related PR: #2446 Contributor: johnlanni Change Log: Update version number to 2.1.5-rc.1 and synchronize the corresponding version information in related files, including Makefile, VERSION file, and Helm charts. Feature Value: This PR primarily updates the project's version information, ensuring that all related configuration files and documentation reflect the latest version number, providing users with accurate version tracking information.
Related PR: #2433 Contributor: johnlanni Change Log: This PR adds release notes in both English and Chinese for version 2.1.4 and updates the licensing configuration to exclude the release-notes directory. Feature Value: By providing detailed release notes, users can better understand the new features and fixes in the latest version, making it easier to adopt and use new software features.
Related PR: #2418 Contributor: xuruidong Change Log: Fixed a broken link issue in the mcp-servers README_zh.md file, ensuring the correctness and usability of document links. Feature Value: By correcting the errors in document links, it enhances user experience when reading and using documentation, avoiding information barriers caused by invalid links.
Related PR: #2327 Contributor: hourmoneys Change Log: This PR mainly updates the documentation related to mcp-server, including adjustments to README_ZH.md and mcp-server.yaml configuration files. Feature Value: By updating the documentation, users can more clearly understand and use the mcp-shebao-tools, providing detailed instructions and configuration examples, enhancing user experience.

📊 Release Statistics

🚀 New Features: 19 items
🐛 Bug Fixes: 14 items
♻️ Refactoring: 2 items
📚 Documentation Updates: 6 items

Total: 41 changes (including 2 significant updates)

Thanks to all contributors for their hard work!🎉

Higress Console Update

📋 Overview of This Release

This release includes 8 updates, covering various aspects such as feature enhancements, bug fixes, performance optimizations, and more.

Distribution of Updates

New Features: 5 items
Bug Fixes: 2 items
Testing Improvements: 1 item

⭐ Key Focus

This release includes 1 significant update that is recommended for close attention:

Feature/issue 514 mcp server manage (#530): The new mcp server console management feature allows users to more conveniently manage and configure the mcp server through the interface, enhancing user experience and operational efficiency.

For detailed information, please refer to the important features section below.

🌟 Detailed Description of Important Features

Here are detailed explanations of the important features and improvements in this release:

1. Feature/issue 514 mcp server manage

Related PR: #530 | Contributor: Thomas-Eliot

Background

In modern microservice architectures, the mcp server, as a key component, is responsible for managing communication between services. However, the existing management systems lack centralized management and visual operations for the mcp server, leading to operational personnel needing to manually configure and manage these services, resulting in inefficiencies and increased error rates. To address this issue, the new mcp server console management feature enables users to easily create, update, delete, and query mcp server instances through a graphical interface. This feature primarily targets system administrators and operational personnel, aiming to enhance their work efficiency and reduce errors.

Feature Details

This change mainly implements the following features:

Create mcp server: Users can create new mcp server instances by filling in necessary parameters in the console interface.
Update mcp server: Users can modify the configuration information of existing mcp servers and save it through the console interface.
Delete mcp server: Users can select the mcp server instances to delete through the console interface and execute the deletion operation.
Query mcp server: Users can query all mcp server instances and their details.

In terms of technical implementation, RESTful API interfaces have been built primarily using the Spring Boot framework, and Swagger has been used to generate API documentation. A new McpServerController class has been added to handle HTTP requests related to the mcp server. Additionally, the Dockerfile has been modified to include copying and permission settings for mcp-related tools. Furthermore, configurations for the SDK have been adjusted to support the new feature.

Usage

To enable and configure this feature, the steps are as follows:

Start the Application: Ensure the Higress Console application is correctly deployed and running.
Access the Console: Access the Higress Console URL through a browser to enter the console interface.
Create mcp server: In the console, select the "mcp server" tab, click the "Create" button, fill in the necessary parameters (such as name, type, etc.), and then click the "Save" button.
Update mcp server: Find the instance to update in the mcp server list, click the "Edit" button, modify the relevant information, and click the "Save" button.
Delete mcp server: Find the instance to delete in the mcp server list, click the "Delete" button, and confirm the deletion operation.
Query mcp server: View all instances and their details in the mcp server list. Note: Before performing any operation, ensure data backup to avoid data loss due to misoperation.

Feature Value

By adding the mcp server console management feature, users can more easily manage and configure mcp server instances, significantly enhancing the system's usability and maintainability. Specifically, this feature brings the following benefits:

Increased Efficiency: Users no longer need to manually write configuration files or perform complex command line operations; management of mcp servers can be completed through a simple graphical interface.
Reduced Error Rate: The visual operation interface reduces problems caused by manual configuration errors.
Enhanced User Experience: The intuitive operation interface enables users to quickly get started, lowering the learning curve.
Increased System Stability: Unified console management ensures consistency and standardization of configurations, reducing stability issues caused by inconsistent configurations.

📝 Complete Change Log

🚀 New Features (Features)

Related PR: #540 Contributor: CH3CHO Change Log: This PR adds a new LLM provider type: vertex to the system by extending the LlmProviderType enumeration and adding the VertexLlmProviderHandler class to support the new provider. Feature Value: The addition of support for vertex as an LLM provider allows users to leverage services provided by vertex, enriching the system's feature set to meet more diverse requirements.
Related PR: #538 Contributor: zhangjingcn Change Log: This PR introduces errorResponseTemplate support to the mcp-server plugin, allowing users to customize error response templates and correcting descriptions in the documentation regarding the conditions for triggering error responses and GJSON path escaping. Feature Value: By providing custom error response capabilities, it enhances user experience and flexibility, allowing developers to adjust how error messages are displayed based on actual needs, thereby better controlling application behavior.
Related PR: #529 Contributor: CH3CHO Change Log: This PR adds the capability to configure multiple model mapping rules for AI routing upstream through adding a pop-up dialog for advanced configuration editing. Feature Value: Users can more flexibly manage the model mappings of AI services, improving configuration efficiency and flexibility, meeting diverse needs.
Related PR: #528 Contributor: cr7258 Change Log: Changed the default PVC access mode from ReadWriteMany to ReadWriteOnce, making it more suitable for most default setting cases. Feature Value: This change reduces unnecessary complexity and improves resource utilization efficiency while providing flexibility for users needing multiple replicas.

🐛 Bug Fixes (Bug Fixes)

Related PR: #537 Contributor: CH3CHO Change Log: Replaced URL.parse with new URL() to address compatibility issues in old browser versions. Feature Value: Improved application compatibility across different browser versions, ensuring a wider user community can use the relevant features normally.
Related PR: #525 Contributor: NorthernBob Change Log: This PR corrects a spelling error in the configuration file, changing 'UrlPattern' to 'urlPattern', ensuring variable naming consistency. Feature Value: By correcting spelling errors, it ensures the accuracy and consistency of configuration files, avoiding service configuration issues caused by case sensitivity, improving system stability and user experience.

🧪 Testing Improvements (Testing)

Related PR: #526 Contributor: CH3CHO Change Log: This PR adds a unit test case to check if the Wasm plugin image is the latest version. It does this by comparing the currently used image tag with the latest image tag's manifest. Feature Value: This feature ensures that the Wasm plugin uses the latest available image, thus enhancing system stability and security, avoiding vulnerabilities or other issues caused by outdated images.