v1.10.3-lts

New features

• Application: Added MCP call node, which has the ability to act as a remote function library service. Users can select tools from the tool set provided by the MCP service for calling; by @liuruibin in #2665
• Application: Added tool calling function to AI dialogue node. After users enable and complete the configuration of MCP tool in tool settings, the large language model will autonomously call the most appropriate MCP tool to handle tasks based on the questions raised by users; by @liuruibin in #2665
• Application: Embed third-party functions to support mobile mode; by @wangdan-fit2cloud in #2615
• Application (X-Pack): In the display settings, support setting whether to display user question avatars and AI reply avatars;
• Function Library: Added built-in function library, including Google search, Bocha search, LangSearch, MySQL query and PostgreSQL query; by @liuruibin in #2611, @wangdan-fit2cloud in #2725
• Function Library: Added startup parameter configuration for functions; by @liuruibin in #2596
• Function Library: Support setting the function logo; by @wangdan-fit2cloud in #2573
• Knowledge Base (X-Pack): Support docking with Feishu Knowledge Base;
• System Settings (X-Pack): Added operation log function;
• System Settings (X-Pack): Enterprise WeChat, Feishu, and DingTalk support password-free login.

Enhancements

• Knowledge base: Added a new operation of generating questions in the knowledge base panel operation to facilitate batch operation; by @shaohuzhang1 in #2760
• Knowledge base: When uploading files in the knowledge base, the segment preview is optimized to lazy loading mode to reduce user waiting time;
• Knowledge base: Optimize the search logic of full-text search and improve the segment recall rate; by @shaohuzhang1 in #2767
• Q&A page: The voice input in the embedded mode of the mobile terminal supports pressing and speaking, and supports releasing and sending after setting automatic sending; by @wangdan-fit2cloud in #2651
• Q&A page: Optimize the voice playback logic, and support streaming playback during automatic playback; @shaohuzhang1 in #2661
• Q&A page: Optimize the interaction of user input parameters. You need to fill in the user input parameters before entering the dialogue. After entering the dialogue page, the user input parameter settings are adjusted to the upper right corner of the question input box; by @wangdan-fit2cloud in #2680
• Q&A page: Optimize the problem that the user feedback operation button displays a line break after the mobile AI reply is completed; by @wangdan-fit2cloud in #2772
• Q&A page: After the AI ​​reply is completed, the form data is filtered out when the user copies the answer; by @shaohuzhang1 in #2778
• Function library: Optimize the location of function permission settings and adjust them to the drop-down list of function panel operations; by @wangdan-fit2cloud in https://github.com/1Panel-
• Application: In the parameter configuration of the form collection node, support adjusting the parameter order; by @wangdan-fit2cloud in #2533
• Application: When a user is having a conversation, if the asker parameter is carried in the URL address, the parameter will be displayed as the questioning user in the conversation log; by @shaohuzhang1 in #2667
• Application (X-Pack): After a user asks a question in WeChat Enterprise, Feishu, or DingTalk, if no reply is received for more than 10 seconds, the system automatically replies: "Thinking, please wait";
• API Documentation (X-Pack): Accessing the Swagger document requires user identity authentication.

Bug Fixes

• Knowledge base: Fixed the problem that the knowledge base name is not fully displayed when selecting the knowledge base page when migrating documents; by @wangdan-fit2cloud in #2718
• Knowledge base: Fixed the problem that the knowledge base name is not fully displayed when switching knowledge bases in the navigation bar inside the knowledge base; by @wangdan-fit2cloud in #2718
• Knowledge base: Fixed the problem that when the uploaded document contains an image with a title set, the image cannot be recognized; by @shaohuzhang1 in #2683
• Knowledge base: Fixed the problem that the document name after uploading a ZIP file to the knowledge base is displayed in garbled characters after decompression; by @shaohuzhang1 in #2747
• Knowledge base: Fixed the issue that when the document name is very long, the document name will be changed to "Sheet" when exporting the knowledge base; by @shaohuzhang1 in #2740
• Application: Fixed the issue that after the child application embedded in the workflow is deleted, the parent application will report an error when opening; by @shaohuzhang1 in #2743
• Application: Fixed the issue that the knowledge base search results are not output in descending order according to the similarity; by @shaohuzhang1 in #2791
• Application: Fixed the issue that the node name suffix is ​​repeated in the workflow, causing parsing errors; by @shaohuzhang1 in #2513
• Application: Fixed the issue that when adding functions with different names in the workflow, the suffix number will be automatically added to the node name; by @shaohuzhang1 in #2513
• Application: Fix the error in setting integer value in form collection slider component; by @shaohuzhang1 in #2601
• Application: Fix the problem that the application avatar is successfully modified, but the prompt information is incorrect; by @wangdan-fit2cloud in #2749
• Application: Fix the problem that when multiple applications are embedded in the same page, users will be prompted that the dialogue does not exist when asking questions in different applications;
• Model settings: Fix the problem of verification error when adding Alibaba Cloud Bailian QWQ type model;
• Login authentication (X-Pack): Fix the problem that after connecting to the OAuth2 platform, the user will return the "500" error code when logging in;
• Login authentication (X-Pack): Fix the problem that multiple OUs (Organization Units) will report errors when connecting to LDAP;
• API Documentation (X-Pack): Fixed an issue where incorrect parameters were output in the OpenAI dialogue interface of the API documentation.