v2.10.8-lts2026-10-22
Security Fixes
■ Security (X-Pack): Fixed a vulnerability where the login authentication configuration endpoint did not redact sensitive fields, exposing third-party login app credentials (such as DingTalk appkey, corpid, and app_secret) in plaintext, which could lead to enterprise app takeover.
Bug Fixes
■ Knowledge Base: Fixed an issue where special characters such as \, /, :, *, and ? could be entered when quickly creating a blank document.
■ Knowledge Base: Fixed an issue where users with only the workspace administrator role were redirected to the agent list page when clicking Import Document in a Feishu knowledge base.
■ Knowledge Base: Fixed an issue where headings with the same content but different levels in a Markdown file were incorrectly recognized during segmentation.
■ Knowledge Base: Fixed an issue where deleting a document after uploading it to a general knowledge base by a custom role returned an error indicating no permission.
■ Knowledge Base: Fixed an issue where the document tag retrieval node passed query parameters when retrieving the tags of the selected knowledge bases, causing the length limit to be exceeded when there were many knowledge bases (#7199).
■ Knowledge Base: Fixed an issue where an empty document list output by the data source tool caused the write to the knowledge base to fail.
■ Agent: Fixed an issue where the button for generating the system prompt was not automatically activated when the default model was used in an AI conversation node.
■ Agent (X-Pack): Fixed an issue where a regular user was prompted with no permission when clicking Save after creating a new agent that referenced a shared tool (#7360).
■ Agent: Fixed an issue where the page title did not display the agent name after refreshing the agent conversation page.
■ Agent (X-Pack): Fixed an issue where the shared tool list incorrectly displayed MCP and Skills items that should be filtered out when adding a tool component in a workflow (#7361).
■ Shared Tool (X-Pack): Fixed an issue where a regular user could not call a shared tool at workflow runtime after adding it to a workflow knowledge base.
■ Resource Management (X-Pack): Fixed an issue where the system administrator failed to import documents into a workflow knowledge base in Resource Management.
■ Resource Management (X-Pack): Fixed an issue where shared tools that had not been granted workspace authorization could be added when editing an agent under a workspace in Resource Management.
■ Resource Management (X-Pack): Fixed an issue where the system administrator failed to download the original document of a knowledge base and was prompted with no permission.
■ Resource Management (X-Pack): Fixed an issue where the document creator was empty when the system administrator created a document for a knowledge base in Resource Management.
■ Resource Management (X-Pack): Fixed an issue where the system administrator failed to view the release history of an agent.
v2.10.7-lts2026-10-15
Enhancements
- Knowledge Base: Special characters such as \, /, ?, *, and [] are now disallowed when renaming documents.
- Agent: A confirmation prompt is now shown when deleting conversation history, to prevent accidental deletion.
- Q&A Page: Tables in AI responses now support horizontal scrolling when the number of columns exceeds the visible width.
Bug Fixes
- Agent: Fixed an issue where content wrapped in html_render tags was rendered blank when exported to PDF or image (#7222).
- Agent: Fixed an issue where long-term memory could be saved without specifying a trigger interval in scheduled trigger mode.
- Agent: Fixed incorrect field name labels in the long-term memory settings panel.
- Agent: Fixed an issue where, when adding agents created at nearly the same time as skills in an AI conversation node, only one of them took effect.
- Agent: Fixed an issue where the execution details of workflow tools were displayed in the wrong order in conversation records (#7221).
- Agent: Fixed an issue where using the long-form writing agent template from Template Center resulted in a "conversation does not exist" error during chat.
- Agent: Fixed an issue where clicking the like or dislike button below an answer caused a system error.
- Agent: Fixed an issue where, after selecting a tool in an MCP node, parameters not marked as required were all treated as required.
- Agent: Visibility rules for user input parameters — optional parameters now filter out unavailable parameters from global variables (#7224).
- Trigger: Fixed an issue where scheduled trigger tasks did not execute and a backend error occurred when saving.
- Knowledge Base: Fixed an issue where, after importing a PDF containing images, the image links were parsed but the images were not displayed.
- Knowledge Base: Fixed an issue where importing large documents into a general knowledge base failed (#7062).
- Knowledge Base: Fixed an issue where the tools bundled with a Template Center template became unavailable after the template was applied in a workflow knowledge base (#7302).
- Knowledge Base: Fixed an issue where full-text search only returned content that exactly matched the keyword (#7234).
Security Fixes
- Security: Fixed a vulnerability where the signing key for Celery task messages could be derived, leading to remote code execution (NVDB-CAIVD-2026894112).
- Security: Fixed a missing function-level authorization (BFLA) vulnerability in the admin oss/get_url endpoint, which could lead to unauthorized access and SSRF (NVDB-CAIVD-2026080673).
- Security: Fixed a vulnerability where chatrecordid could be used to overwrite other visitors' conversation records (session integrity).
- Security: Fixed a vulnerability where the like and dislike endpoints still accepted cross-client operations.
- Security: Fixed an Eval injection vulnerability in the "Init Param" validation rules of custom tools, which allowed arbitrary JavaScript execution.
- Security: Fixed a path traversal vulnerability in knowledge base names that allowed writing arbitrary .xlsx files and poisoning instance-wide import templates.
- Security: Fixed an account takeover vulnerability.
- Security: Fixed an unauthenticated SSRF vulnerability in the chat file proxy caused by an incomplete host denylist.
- Security: Fixed a vulnerability where the MCP configuration URL was not validated, allowing users with the USER role to perform SSRF against arbitrary addresses and read back responses (direct connection from the web process, bypassing the sandbox).
- Security: Fixed a vulnerability where the login CAPTCHA was not destroyed after successful validation, enabling unauthenticated online password brute-forcing (admin takeover).
- Security: Fixed a vulnerability where authenticated users could access cross-tool execution records through an unscoped cache hit.
v2.10.6-lts2026-09-03
Security Vulnerability Fixes
- Security: Fixed a security vulnerability where Chat file uploads could spoof APPLICATION attribution and be read anonymously, allowing content injection into the application namespace.
- Security: Fixed a security vulnerability where document/paragraph migration only validated source knowledge base permissions, leading to cross-tenant content injection and RAG poisoning.
- Security: Fixed a security vulnerability where XLSX embedded image decompression lacked a pixel count limit, allowing authenticated users to trigger memory-exhaustion denial of service (DoS).
- Security: Fixed a security vulnerability where problem/batch_association lacked ownership validation, leading to cross-tenant paragraph content disclosure (IDOR).
- Security: Fixed a security vulnerability where the anonymous application profile endpoint over-serialized data, causing sensitive configuration disclosure (/chat/api/application/profile).
- Security: Fixed a security vulnerability where the password reset verification code endpoint lacked a guessing attempt limit, allowing unauthorized account takeover.
- Security: Fixed a security vulnerability where the function library sandbox could be bypassed (dlopen stack check bypassable + dlsym(RTLD_NEXT) not intercepted) (CVE-2026-79919).
- Security: Fixed a security vulnerability where Python's support for file descriptor form os.execve(fd,...) could be leveraged to enter the libc fexecve call path, causing sandbox bypass (NVDB-CAIVD-2026208648).
- Security: Fixed a security vulnerability where guest session ownership validation was missing, leading to cross-guest-session writes (Session Integrity Violation).
- Security: Fixed a security vulnerability where the default hardcoded Django SECRET_KEY allowed signed tokens to be forged offline, resulting in unauthorized file access.
New Features
- Workflow Default Model Settings
In the MaxKB v2.10.6 LTS community edition, workflow default model settings have been introduced for agent workflows, knowledge base workflows, and workflow tools. Previously, multi-AI-node workflows required opening each node individually to modify model configurations, which was relatively cumbersome. Now, a single configuration takes effect globally, significantly reducing maintenance costs and improving the consistency and efficiency of batch model switching. Core features of the default model settings include:
- A workflow-level default model configuration entry;
- Newly created nodes automatically inherit the default model;
- Nodes support flexible model selection modes (Default / Custom / Reference Variable, depending on node type);
- The default model can be applied to all nodes with one click, uniformly switching all nodes' model sources to the default model;
- A single change to the default model takes effect across multiple nodes;
- Configuration completeness is ensured before publishing.
- Tencent Hunyuan & Tencent Cloud Providers Merged into the New "Tencent Cloud" Provider (Integrated with Tencent Cloud TokenHub)
Due to adjustments by Tencent Cloud, in the MaxKB v2.10.6 LTS community edition, the Tencent Hunyuan and Tencent Cloud providers have been merged into a new "Tencent Cloud" provider. All models previously configured based on the Tencent Hunyuan or legacy Tencent Cloud provider integration will no longer be usable after the upgrade. After upgrading, please go to the "Model Providers" page, select "Tencent Cloud" from the provider list, and reconfigure your model access information according to the TokenHub integration requirements. Once configured, the models will be available again.
Feature Optimizations
- Knowledge Base: The knowledge base/document list selector now supports search when migrating documents/paragraphs.
- Agent: The knowledge base component in the user input area now supports search.
- Agent: In the "Add to Knowledge Base" feature on the conversation log page, the knowledge base and document list selectors now support search.
Bug Fixes
- Knowledge Base: Fixed an issue where importing a CSV file with over-sized cell content resulted in empty reads and a "field larger than field limit (131072)" error in the backend (#6756).
- Knowledge Base: Fixed an issue where importing Excel spreadsheets without recognizing data boundaries caused large numbers of empty columns to be imported, leading to import timeouts (#6355).
- Knowledge Base: Fixed an issue where entering 33 consecutive "-" characters in full-text search during hit testing caused a search error.
- Agent: Fixed an issue where conversations errored when output content contained null characters (\u0000, \x00, etc.) (#6722).
- Agent: Fixed an issue where AI nodes in advanced agents could not receive data passed from previous nodes when using the Ascend MindIE large model (#6681).
- Agent: Fixed an issue where calling the model errored when the "Role Setting" of an advanced orchestration AI node was empty (#6662).
- Agent: Fixed an issue where, after invoking a tool, the execution source shown in tool execution records displayed the tool name instead of the agent name (#6647).
- Agent: Fixed an issue where files uploaded to workflow knowledge bases were temporary files, causing preview and download failures after the files were cleaned up (#6785).
- Agent: Fixed an issue where an agent used as a skill of an AI conversation node could not pass parameters other than message.
- Agent: Fixed an issue in advanced agents where the default knowledge base could not be set after changing the component type of user input parameters to knowledge base.
- Agent: Fixed an issue where an error occurred when clearing chat logs (#6690).
- Model: Fixed an issue where the MiniMax H3 text-to-video model reported errors during use after successful integration.
v1.10.15-lts2026-08-11
Bug Fixes
- Security Vulnerability: Fixed flaws in permission verification to block a high-risk privilege escalation vulnerability where anonymous application session tokens could be exploited to obtain the administrator's system API Key.
- Knowledge Base: Fixed garbled paragraph content when uploading PDF files where the title text is identical to the body content.
v2.10.5-lts2026-08-06
Security Vulnerability Fixes
- Security Vulnerability: Fixed missing‑owner‑check at chat‑sharing link endpoint, where chat tokens could be abused to share other users’ conversation content.
- Security Vulnerability: Fixed privilege escalation allowing regular users to fetch user information of other workspaces via API (#6576).
- Security Vulnerability: Fixed privilege escalation where any user’s API Key could be used to delete arbitrary files.
- Security Vulnerability: Fixed the issue that regular users could obtain sensitive information such as other users’ email addresses via API (#6577).
Bug Fixes
- Knowledge Base: Fixed failure‑to‑load issue for images returned from knowledge bases within conversations.
- Knowledge Base: Fixed abnormal paragraph content when uploaded PDF contains identical text in titles and body content (#6543).
- Knowledge Base: Fixed incorrect display of the “Allow preview in knowledge source” option in the web‑knowledge‑base import‑document modal.
- Agent: Fixed truncated content when copying long conversation text (#6568).
- Agent: Fixed inaccessible AI‑generated images when agent is embedded in third‑party pages.
- Agent: Fixed inaccessible images generated by the agent image‑generation node.
- Agent: Fixed error popup after agent debugging completes, which blocked access to execution details.
- Agent: Fixed basic agents being able to run retrieval against unpublished knowledge‑base retrieval configurations (#6519).
- Agent: Fixed image‑loading failure after images are extracted by the file‑content‑extraction node.
- Agent: Fixed execution failures of agent trigger tasks.
- Agent: Fixed workflow‑tool‑node outputs not being written into conversation logs.
- Agent: Fixed failure of the sendMessage method in agent opening remarks to pass user parameters, causing loss of user questions.
- Agent: Fixed file preview/download failures caused by insufficient permissions in knowledge‑source panel on Q&A page.
- Agent: Fixed permission‑denied prompts for partial files when downloading multiple uploaded files from conversation logs.
- Agent: Fixed inaccurate active‑user statistics on overview page caused by residual uncleaned data after conversation‑log cleanup.
- Role (X‑Pack): Fixed missing permission‑bit controls for certain functions.
- Shared Resources (X‑Pack): Fixed wrong page redirection when system administrators click “Upload Document” inside shared knowledge bases.
- Shared Resources (X‑Pack): Fixed access to hit‑test results for shared knowledge bases within workspaces.
v2.10.4-lts2026-07-16
Security Vulnerability Fixes
- Vulnerability: Fixed a system file permission flaw that allowed anonymous access to resources;
- Vulnerability: Fixed an IDOR vulnerability in admin chat logs which allowed bypassing application path authentication to read cross-application chat records;
- Vulnerability: Added independent authorization verification mechanism for agent and workflow tool scheduling;
- Vulnerability: Fixed missing host validation when UpdateStoreTool fetches external app store URLs;
- Vulnerability: Fixed the horizontal privilege escalation issue that enabled users to access and modify unauthorized documents and segments across knowledge bases;
- Vulnerability: Fixed expired agent API keys being usable at the /chat/api/mcp endpoint;
- Vulnerability: Fixed horizontal privilege escalation caused by defective MCP tool authorization logic;
- Vulnerability: Fixed exposure of application IDs on the homepage ranking page, which could lead to unauthorized calls to other users' application nodes;
- Vulnerability: Fixed the risk of cross-workspace privilege escalation when viewing and editing model parameter forms;
- Vulnerability: Fixed prompt injection vulnerability to mitigate the risk of agent command execution;
- Vulnerability: Improved internal IP validation for web document import and sync crawlers to remediate SSRF risks.
Bug Fixes
- Agent: Fixed the issue that newly added vision models for image understanding nodes in advanced agents were not displayed in the model list in real time (#6402);
- Agent: Fixed incorrect prompt messages during conversations after deleting the LLM bound to basic agents;
- Agent: Fixed failures when agents call shared models with the error "Model does not exist";
- Knowledge Base: Fixed misclassification of segmented text as table content when there is no blank line between tables and subsequent content in Word files (#6396);
- Knowledge Base: Fixed cross-document contamination of segment position data when dragging and reordering document segments (#6373);
- Homepage: Fixed the issue where agents without user permissions could be searched by name on the ranking detail page;
- Trigger: Fixed duplicated task display in task details of scheduled trigger lists;
- Trigger: Fixed missing default value passing for user input parameters of advanced agents under trigger scenarios, resulting in empty parameter values (#6320);
- Trigger: Fixed parameter passing errors when agent user inputs include toggle-type parameters in triggers;
- User Management: Fixed the issue that the user listing API only returns the first 200 users and cannot fetch all user records (#6375).
v2.10.3-lts2026-07-03
Feature Optimizations
- Knowledge Base: Added visual progress display for file uploads; files that failed to upload support one-click re-upload.
- Agent: Supported batch configuration for agent conversation log cleanup strategies (#6213).
- Agent (X-Pack): Enabled voice input capability for WeCom robot integration.
- Agent: Optimized copywriting on the overview page to align its visual style with the homepage.
- Role Permissions (X-Pack): Added homepage view permission configuration for regular users and workspace administrators.
Bug Fixes
- Agent (X-Pack): Fixed the 500 error in conversations caused by multiple agents sharing the same chatsessionid when integrated with WeCom (#6325).
- Agent: Fixed the issue where variables for interface parameters in advanced agents could not be assigned values during debugging (#6271).
- Agent (X-Pack): Fixed the UI issue where the mobile input box covers the like and dislike buttons after enabling the disclaimer (#6249).
- Agent: Fixed invalid visibility control logic for external user input parameters.
- Q&A Page: Fixed abnormal style rendering of video content in replies on mobile devices.
- Knowledge Base: Fixed parsing errors when uploading PDF files containing special characters.
- Knowledge Base: Fixed the issue that video files stored in the knowledge base were not exported during knowledge base export.
- Knowledge Base: Fixed permission errors that prevented regular users from loading the tool list via MCP call nodes in workflow knowledge bases.
- Knowledge Base: Fixed missing reference links for tools inside loop blocks after importing workflow knowledge bases, which made tool nodes unavailable.
- Knowledge Base: Fixed occasional duplicate display of segmented content in the segment preview after document upload.
- Models: Fixed errors when invoking the Qwen3-VL-Reranker-8B reranking model from the vLLM provider in multi-path recall nodes (#6321).
- System Settings (X-Pack): Fixed the brief flash of the default MaxKB Logo after page refresh when a custom Logo is configured (#6295).
v2.10.2-lts2026-06-18
Feature Optimizations
- Q&A Page: Optimized voice input logic to append transcribed text instead of overwriting existing input content (#6212);
- Q&A Page: Added a clear input content function after users type content in the input box (#6212);
- Q&A Page: Users can normally send new questions after switching to Session 2 while Session 1 is generating replies;
- Agent: Files submitted via forms can be downloaded from conversation logs (#5271);
- Agent: The sendMessage method in the html_render tag of advanced agent opening remarks supports passing API parameters and user input parameters;
- Agent: Added tool source information display in the operation menu of tool nodes for advanced agents;
- Agent (X-Pack): Added permission control configuration for conversation sharing in display settings;
- Agent (X-Pack): Input "/new" to automatically create a brand-new conversation for third-party application integrations;
- Knowledge Base: Columns in the document list support manual width adjustment;
- Knowledge Base: Added a creator field to the document list, and the creator of newly uploaded documents will be recorded (#5240);
- Knowledge Base: Provided two export modes allowing users to choose whether to export original source files when exporting a knowledge base (#6137);
- Knowledge Base: Optimized pagination logic for tag management to prevent frontend lag caused by excessive data under a single tag;
- Models: Added support for the Imagen image generation model under the Gemini provider (#5234);
- Models: Added support for the VEO video generation model under the Gemini provider (#5234);
- Models: Updated the model dropdown list of all model providers to show the latest models;
- Permission Management: Users who only have tool view permission are prohibited from accessing the tool editing page;
- Workspace (X-Pack): Optimized lazy loading performance of the workspace list.
Bug Fixes
- Q&A Page: Fixed the issue that the display of node execution progress was not controlled by the execution detail toggle switch;
- Q&A Page: Hide node execution progress when AI reply is paused, and resume displaying progress after clicking Continue;
- Agent (X-Pack): Fixed the failure to preview PDF files from knowledge sources in third-party integrated conversations (#6202);
- Tools: Fixed the missing required field marker for the name input box when creating a new MCP tool;
- Knowledge Base: Fixed corrupted downloaded DOCX source files after upgrading from v2.3.1 to v2.10.1 (#6238);
- Knowledge Base: Fixed execution errors thrown by the Write-to-Knowledge-Base node during workflow debugging;
- Models: Fixed errors when adding the qwen3-tts-flash model from Alibaba Cloud Bailian provider (#6236);
- Login Settings: Fixed the vulnerability allowing Community Edition users to log in without captcha verification (#6214).
v2.10.1-lts2026-06-10
Bug Fixes
- Knowledge Base: Fixed the issue where the QA pair template content was displayed incorrectly when downloaded.
- Agent: Fixed the issue where questions could not be sent during debugging for agents with user input parameters.
- Agent: Fixed the required field error when directly asking a question with the default value kept for an external toggle parameter whose default value is false.
- Agent: Fixed the issue where clicking the "More Parameters" button during debugging/conversation affected the state of external toggle parameters.
- Agent: Fixed the issue where the Skill execution directory could be randomly cleared under concurrent scenarios (#6185).
- Agent: Fixed the issue where variables referenced in the system prompts of Image Understanding and Video Understanding nodes were not resolved (#6180).
- Tools: Fixed the dependency not found error when using pymysql in Python scripts (#6182).
v1.10.14-lts2026-06-08
Bug Fixes
- Models: Fixed the error when adding reranking models from Alibaba Cloud Bailian.