v3.0.12026-11-10
BISHENG v3.0.1 — patch release
Bug fixes
- Workflow input / workspace chat: plain-text Word files failed to parse. Uploading a .docx that contains no images (via the Input node's dialog or form input, or in workspace chat) failed with [Errno 2] No such file or directory: '.../media', and only worked after adding an image to the document. Pandoc creates the media directory only when a document contains images; the image-reference rewrite step listed that directory without checking it exists. Regression introduced in v3.0.0. (7df786081d)
Upgrade
Change the image tags in docker/.env from v3.0.0 to v3.0.1 (backend, frontend, sandbox, runtime-manager, app-proxy), then docker compose pull && docker compose up -d. No database migration.
Full Changelog: https://github.com/dataelement/bisheng/compare/v3.0.0...v3.0.1
v3.0.0:Vibe App Hosting & Integration with DeepSeek Harness2026-11-04
<img width="1914" height="956" alt="image" src="https://github.com/user-attachments/assets/7248f659-9fa5-4df2-91ab-87e9fcef7021" />
🚀 BISHENG 3.0 Release:Vibe App Hosting & Integration with DeepSeek Harness
🏭 1. Vibe App Hosting
> Overview: Developers can host apps they build locally on the BISHENG platform. After an app is approved and goes live, users with access can open and use it on the platform. This feature is off by default. To turn it on, the ops team enables it in config.yaml and deploys the app runtime components.
1.1 💻 Local development and deployment: Developers install the bisheng command-line tool and sign in with a service account key. They can then develop and debug apps locally, declare the runtime, resource tier, and the platform models and knowledge bases the app uses in the app manifest, and submit the app to the platform.
1.2 ✅ App approval and preview: Approvers see app publishing requests in the Approval Center. They can browse the source file tree and preview the app, then approve or reject the request.
1.3 🛠️ Hosted app management: App owners open a hosted app under Build → Apps. Owners can view the current resource tier, set who can see the app, enter third-party keys, view and edit app data online, and bring the app online, take it offline, or delete it.
1.4 🏪 Using apps: Once an app is live, employees with access can find it in the app square. They click Enter app to open it in a new tab and use it right away, with no second sign-in.
1.5 🔌 External interfaces
- MCP Server: Capabilities already in the open API, including knowledge base management and retrieval, organization and identity lookup, and hosted app status and data, are now also available over the MCP protocol. Any MCP-compatible agent can call these tools.
- OpenAI-compatible model API: Hosted apps can call the models configured on the platform.
- bisheng-sdk: Hosted apps can use the SDK to access platform identity, knowledge base retrieval, and attachment storage.
🖥️ 2. Workspace Redesign
2.1 🧭 Global sidebar: The desktop workspace now uses a global sidebar on the left. From top to bottom it shows New chat, Knowledge Space, Subscription, Apps, More, the chat list, and user info.
2.2 🏬 Discover: Discover is added under More in the sidebar. Users can browse knowledge spaces, channels, apps, and skills on one page, organized by tabs.
2.3 🧩 Tools & Skills center: Tools & Skills is added under More in the sidebar. Users can browse tools and skills by category, and upload their own skills under My skills. Admins can grant built-in tools and enterprise skills to specific people or departments under Build → Tools & Skills.
2.4 🎙️ Streaming voice input: After a user clicks the voice button in the input box, the transcript appears as they speak. Real-time speech recognition models from Tongyi (Alibaba), Volcengine, Baidu, and Tencent are supported.
🧠 3. Deep Research and Tools
3.1 📈 Deep Research quality improvements: The PPT skill adds 5 built-in templates, and generated decks can be previewed online. A new web report skill improves HTML output. Excel file handling is improved. History is compressed and long tool results are offloaded to files automatically, based on the context window of the selected model. Thinking models no longer redo their reasoning as often across multiple calls. Long tasks no longer freeze the browser.
3.2 📦 Container code execution sandbox: The code executor adds a container sandbox mode. Code runs in a separate container, isolated per session. New installations use the sandbox by default. Upgraded environments keep their current mode; admins can switch by selecting Run in sandbox for the code executor under Build → Tools.
3.3 🌐 DeepSeek web search and web fetch tool: The web search tool adds DeepSeek as a search engine. A new web fetch tool lets agents read the main text of a web page from its URL. Daily mode, Deep Research, assistants, and workflows can all use these tools.
⚙️ 4. Other Improvements
- MinerU 4.0 parsing engine support: Knowledge bases can use MinerU 4.0 to parse PDFs, images, and OFD files.
- Fixed continuously growing memory usage of Deep Research processes on DM (Dameng) databases.
💡 5. Upgrade Notes (see >> Upgrade Notes)
- This upgrade requires downtime, plus a database and permission model upgrade.
- Local code execution now runs in isolation by default. Tasks that need a lot of memory may be affected.
- Vibe app hosting and the container code execution sandbox are off by default. To use them, follow the upgrade guide.
v2.6.0-fix22026-08-11
Full Changelog: https://github.com/dataelement/bisheng/compare/v2.6.0-fix...v2.6.0-fix2
What's Changed
* backport(linsight): 把 3.0-beta1 的任务模式改进回合到 2.6 by @jieyuhuayang in https://github.com/dataelement/bisheng/pull/2263
* fix(linsight): 撤掉「模型谎报文件名就替它造出来」,改为检测并记录 by @jieyuhuayang in https://github.com/dataelement/bisheng/pull/2265
* feat(linsight): 任务模式产出区的四项后续打磨 by @jieyuhuayang in https://github.com/dataelement/bisheng/pull/2267
Full Changelog: https://github.com/dataelement/bisheng/compare/v2.6.0-fix...v2.6.0-fix2
v2.6.0-fix2026-07-30
What's Changed
fix: correct variable name in replaceparam closure by @hobostay in https://github.com/dataelement/bisheng/pull/2080
* fix: remove stale dev upstream from nginx config by @chenzefeng09 in https://github.com/dataelement/bisheng/pull/2092
* ci(security): avoid script injection from commit message in workflows by @kobihikri in https://github.com/dataelement/bisheng/pull/2200
* Feat/2.6.0 by @zgqgit in https://github.com/dataelement/bisheng/pull/2205
* Update README.md by @wisdomqin in https://github.com/dataelement/bisheng/pull/2206
* Hotfix/2.6.0 by @zgqgit in https://github.com/dataelement/bisheng/pull/2233
New Contributors
* @hobostay made their first contribution in https://github.com/dataelement/bisheng/pull/2080
* @chenzefeng09 made their first contribution in https://github.com/dataelement/bisheng/pull/2092
* @kobihikri made their first contribution in https://github.com/dataelement/bisheng/pull/2200
Full Changelog: https://github.com/dataelement/bisheng/compare/v2.6.0...v2.6.0-fix
v2.6.0 — Linsight Task Mode & Enterprise Permission Upgrade2026-07-18
<img width="1786" height="989" alt="image" src="https://github.com/user-attachments/assets/c069aa98-8679-4578-b5bc-4a670aca25b1" />
> This release consolidates everything from v2.5 and v2.6.
🧠 1. Linsight Upgraded to Task Mode
Linsight has been re-architected and upgraded, adding task planning, sub-agent delegation, a virtual file system, human-in-the-loop (HITL) collaboration, and skill support — for more stable planning, more accurate tool calls, long processes that are less likely to go off track, faster execution, and better results.
1.1 📋 Plans Before It Acts: Task mode breaks your request into a visible to-do list and works it top to bottom, so you always see which step it's on and what's next.
1.2 🤝 Delegates the Deep Digging: Heavier investigation is handed to research sub-agents that work in an isolated context and fold a distilled summary back into the main task.
1.3 💬 Human-in-the-Loop (HITL): When your intent is ambiguous, the task pauses to ask a clarifying question — with interactive options — and resumes automatically once you answer.
1.4 🧩 Skills, On Your Terms: Administrators can create, upload (.md / .zip / .skill), or import skills from GitHub, and enable, disable, or delete them in the console.
1.5 ⏹️ Stop & Resume Across Turns: Halt a running task at any time and continue the same conversation later — subsequent turns can see and build on the files produced earlier.
🔐 2. Enterprise Permission System Upgraded
The permission system is upgraded to a combined RBAC + ReBAC model, letting large organizations model their real structure and share with precision. Full feature details >> https://dataelem.feishu.cn/wiki/BWSMwdc2Ciwif4kfe5HcyGHgn3b#share-QWTRdhvMiofvSex2h81c6IR6nBg
2.1 🌳 Model Your Org as It Really Is: Build a multi-level department tree in the admin console — create, rename, move under a new parent, and delete — with department administrators managing their own department's members and roles, and user groups to gather people from across departments.
2.2 🎚️ Share at Four Relationship Levels: Whether it's a knowledge space, document library, folder and file, workflow or assistant app, tool, dashboard, or channel, permission management lets you grant Owner / Can Manage / Can Edit / Can View to an individual, a department (optionally its whole subtree), or a user group.
2.3 🛡️ Permission Guardrails: A non-super-admin can never grant above the level they hold, every resource always keeps at least one owner, and anyone who gains or loses access is notified in-app. After upgrading from an earlier version, existing role-, group-, and space-member permissions migrate into the new model in one operation.
2.4 🎫 Roles, Menus & Quotas: Policy roles control which feature menus a user sees after login, with three per-role quotas — knowledge-base upload capacity (GB), channels that can be created, and knowledge-space subscriptions.
⚡ 3. A Smarter Workstation with Inline Citations
In everyday chat, the model now judges when to call a tool (including knowledge-base retrieval) and uses citation badges to mark the source of every grounded answer.
3.1 🤖 Agentic Tool Calling: With one or more tools or knowledge bases enabled, the model decides whether and when to call them and can chain several calls within a single answer. Administrators can customize which tools are available in daily mode from the admin console.
3.2 🔎 Inline Source Citations: Across daily mode, workflows, and assistants, answers carry numbered citation badges — blue for knowledge sources, purple for web search. Click one to preview the source (PDFs box-highlight the exact cited passage), or open the References panel under an answer to see every source it drew on.
3.3 📤 Export & Reuse Conversations: Pick the answers you want and export them as Word (.docx), PDF, Markdown (.md), or TXT, or send them straight into a knowledge space — down to a specific folder.
📚 4. A More Powerful Knowledge Space
4.1 🔗 Use Knowledge Spaces in Apps: Assistants and workflow nodes (Knowledge Q&A, Knowledge Retrieval, and Assistant) can now retrieve from knowledge spaces, not just document knowledge bases. An optional switch filters retrieval and citations by the running user's file-level permissions, so a file a user can't see never reaches the answer.
4.2 🌐 Import Web Pages & Audio/Video: Paste a URL to fetch a page's main content as a knowledge file, or upload audio/video to have it transcribed to timestamped text and ingested. Transcription runs through a speech model configured in the console's system model settings.
4.3 🗂️ Document Version Management: Keep every revision of a document under one entry, with similarity hints when you upload a near-duplicate; knowledge-space search always returns only the primary version, and switching the primary takes effect with no re-indexing. Version management is off by default and can be enabled in the system YAML configuration.
4.4 🛡️ Content Security Review: Tenant administrators can turn on sensitive-word screening — independently for knowledge-space uploads and subscribed channel articles — using built-in or custom word lists. A flagged upload is blocked from the index; a flagged channel article is marked "under review" and hidden from ordinary users.
4.5 ✂️ Stronger Ingestion: Subscribed channel content can be filed into a chosen knowledge space automatically; a new "chunk by hierarchy" strategy segments Markdown and Word documents along their own heading outline; and PPT parsing has been tuned to resolve the image and text loss that could occur during PPT parsing.
🇨🇳 5. Ready for Domestic Deployment (Dameng DM8)
To meet the localization requirements of government and enterprise customers, BiSheng now runs its entire metadata layer on Dameng DM8, alongside MySQL with functionality on par.
5.1 🗄️ DM8 as a First-Class Database: Point the database connection at Dameng to deploy the whole platform on DM8 — passwords stay encrypted and the platform handles the dialect differences transparently.
5.2 🔁 Migration & Full-Stack Localization: A MySQL-to-Dameng migration tool moves existing data across and schema upgrades run automatically; in the commercial edition, the Java gateway and the OpenFGA permission engine also support Dameng for end-to-end localized deployment.
✨ 6. More Improvements
- Conversation history on shared app pages: Both login-required and no-login application pages keep a conversation list so users can resume an earlier chat — server-side per account when signed in, browser-local per device when not.
- Mobile-optimized Workstation: Core pages adapt to phones on their own, with no separate app.
- Workflow input-node file strategies: The input node accepts multiple files and lets builders choose how each file is used downstream.
v2.4.0 Enterprise Shared Knowledge Space2026-05-14
<img width="2880" height="1560" alt="image" src="https://github.com/user-attachments/assets/f331da72-b058-45fb-a1ef-b5040352809a" />
📚 1. Workstation Personal Knowledge Base Upgraded to Knowledge Space
The former Workstation personal knowledge base has been upgraded to "Knowledge Space" — an enterprise-grade "IMA" for individuals and teams, ensuring corporate knowledge assets stay secure and never leave the organization.
1.1 📂 File Management + Tag System: Multi-level folders, tag classification, and card/list view toggle for quick knowledge asset organization.
1.2 🔍 Full-Text Search + AI Q&A: Combined keyword-tag search; AI assistant can answer questions against a single document or an entire space.
1.3 👥 Member Management & Collaboration: Invite others to join your knowledge space with three role tiers (Creator / Admin / Member). Admins can upload/delete files, edit tags, and approve join requests for team-level co-management.
1.4 🏪 Knowledge Plaza: Browse public and approval-based knowledge spaces in one place, with search, preview, and one-click join/apply — enabling cross-team knowledge discovery and reuse.
📡 2. Subscriptions — Your Enterprise Intelligence Feed
Aggregate content from WeChat Official Accounts and websites with daily auto-updates, so your team never misses critical information.
2.1 🏷️ Channels + Smart Filtering: Create topic channels (up to 50 sources) with boolean keyword filtering rules and sub-channel support. Channel articles can be saved to Knowledge Space with one click, closing the full "intelligence → knowledge" loop.
2.2 🤖 Article AI Assistant: One-click article summarization and full-text Q&A.
2.3 🏪 Channel Plaza + Sharing: Public channels can be discovered and subscribed to, with link sharing support.
🖥️ 3. Workstation Redesign
Reimagined navigation and interaction, enabling users to complete their daily workflow in a single interface.
3.1 🗂️ Four Top-Level Entries: Home, Apps, Subscriptions, Knowledge Space — clearer navigation with state memory across tabs.
3.2 💼 App Center Upgrade: Pin frequently used apps + continue conversations; each app now has its own dedicated chat interface with shareable app links.
3.3 📊 Enhanced Daily Mode: Mount both Knowledge Spaces and backend organization knowledge bases simultaneously for parallel dual-source retrieval.
3.4 💬 In-App Messaging: Approve channel subscription requests and knowledge space join applications directly from the message center.
⚙️ 4. Other Improvements
4.1 📄 Switchable Paddle OCR / MinerU document parsing engines for self-controlled document processing.
4.2 🔐 Role-based permissions now extend to Workstation menus (Subscription / Knowledge Space independently configurable); audit logs expanded accordingly.
4.3 👤 Avatar upload and password change added to Personal Center.
v2.3.0-release2026-01-28
🚀 BISHENG 2.3-Release
<img width="1957" height="996" alt="image" src="https://github.com/user-attachments/assets/fcf4b379-1121-40a0-91b6-5da4717bdcbd" />
📊 1. Prebuilt Dashboards & Data Insights
- Open-source edition: Comes with standard prebuilt dashboards covering four core metrics: number of users, number of apps, number of knowledge bases, and user feedback.
- Commercial edition (Business Expansion Suite): Provides full dashboard coverage out of the box, supports customizing component dimensions and metrics based on 11 intermediate tables, and enables multi-dashboard management.
🎨 2. Dashboard Management & Personalization (Commercial Feature)
🔄 2.1 Full Lifecycle Management: Supports the full lifecycle of dashboards, including creating, duplicating, renaming, deleting, publishing, sharing, and taking offline. The admin console supports role-based permissions for dashboard management and usage.
✨ 2.2 Drag-and-Drop Layout & Interactions: Allows dashboard components to be freely dragged, resized, and duplicated, and supports full-screen presentation mode.
📈 2.3 Rich Dashboard Components: Built-in components include bar charts (basic/stacked/grouped), horizontal bar charts (basic/stacked/grouped), line charts (basic/stacked/area), pie charts, donut charts, KPI cards, plus a global time filter component, meeting diverse data visualization needs.
🔧 2.4 Advanced Chart Configuration: Supports drag-and-drop configuration of dimensions/metrics, custom number formatting, advanced filtering and sorting, and personalized styling.
🌍 3. Other Improvements
📚 3.1 Knowledge Base File Parsing Improvements: Improved processing logic for docx and ppt files, addressing parsing failures in certain edge cases.
👥 3.2 System Management Improvements: Supports role-based access control for “Workspace” and “Admin Console” spaces, and adds audit logs for dashboard-related actions.
v2.2.02025-11-11
🚀 BISHENG 2.2-release Version
<img width="1695" height="994" alt="image" src="https://github.com/user-attachments/assets/0c5efc99-7100-48ac-bf54-0e4341636fd0" />
🎙️ 1. Workspace supports voice interaction (ASR+TTS):
1.1 Speech input (ASR): Click the “microphone” icon in the input box to dictate text directly.
1.2 Voice playback (TTS): A new “speaker” icon appears next to AI replies; click it to listen to the audio readout.
1.3 Model support: In the admin system’s model settings, you can configure ASR and TTS models. Provider formats from OpenAI, Azure, and 通义千问 are supported.
✨ 2. Conversation experience upgrades:
2.1 Enhanced Markdown rendering: Conversation content now supports rendering Echarts charts and Mermaid flowcharts.
2.2 Conversation sharing: A new “Share” button lets you generate a public, read-only link to safely share conversation results with others.
💡 3. Workflow RAG capability upgrades
3.1 New “Knowledge Base Retrieval” node: A standalone document knowledge base retrieval node for use within workflows, allowing developers to perform retrieval operations more flexibly and obtain structured results (including metadata and text).
3.2 Advanced retrieval settings: “Advanced retrieval settings” are now available in both the “Knowledge Base Retrieval” node and the existing “Document Knowledge Base Q&A” node. You can adjust hybrid retrieval weights (keyword vs. vector), result length, and enable a Rerank (re-ranking) model for secondary ranking.
🎛️ 4. Model management optimizations
4.1 Advanced model parameters: Model management (LLM/Embedding) supports advanced parameter configuration. You can pass vendor-specific parameters (e.g., for 阿里云, 火山引擎) in JSON—such as temperature, enable_thinking—to support configuring hybrid “thinking” model modes and related needs.
4.2 Baidu Qianfan Cloud Platform authentication logic update: Qianfan has switched to an OpenAI-compatible mode (using Base URL and API Key). Users of Baidu models, please update your configuration.
🔐 5. App/Tool admin permissions: Apps (workflows, assistants, skills) and tools now support assigning “admin permissions” to facilitate team collaboration and management.
📚 6. Knowledge base management optimizations: The QA knowledge base now supports “Duplicate” and “Settings” (including changing the embedding model).
🛠️ 7. Historical knowledge base defect fix
7.1 Defect description: When the Document Knowledge Base and the QA Knowledge Base use the same embedding model, modifying content in either may affect the other’s data in Milvus, leading to unexpected retrieval results.
7.2 Root cause: In Milvus’s partition mode, multiple knowledge bases using the same embedding model share one collection. The files in the Document Knowledge Base and QA pairs in the QA Knowledge Base may have conflicting fileid values (different tables with auto-increment IDs). Therefore, operations on items with the same fileid—files or QA pairs—will affect the counterpart’s data in Milvus.
7.3 Fix: ① Upgrade to the 2.2 release; ② Use the script (https://dataelem.feishu.cn/wiki/WNqbwm8zyiPo9ukIw5ecCWcfnPe#share-MTpPdLnTXo0hl3xN7PqcXp6NnBe) to scan for problematic knowledge bases; ③ Run the script as needed to repair all knowledge bases.
v2.1.12025-09-19
What's Changed
* fix: 工作台默认配置 by @zgqgit in https://github.com/dataelement/bisheng/pull/1636
* fix: 工作台保存修改 by @zgqgit in https://github.com/dataelement/bisheng/pull/1637
Full Changelog: https://github.com/dataelement/bisheng/compare/v2.1.0...v2.1.1
v2.1.02025-09-18
BISHENG 2.1 release 版
🚀 1. 工作台正式发布
<img width="1280" height="736" alt="image" src="https://github.com/user-attachments/assets/47c5921d-f7c4-440c-b14a-decc3c7f0550" />
1.1 🏁 BISHENG 工作台独立上线: 定位为面向业务人员的 AI 应用门户;原有 BISHENG 平台将聚焦为管理与构建后台。
1.2 🗂️ 应用中心上线: 集中展示已构建的各类应用,支持按标签分类与设置常用应用。
1.3 📚 个人知识库优化: 1)支持管理员在系统模型设置中切换 Embedding 模型并重建个人知识库;2)个人知识库问答效果优化。
1.4 ⏱️ 系统提示词支持传入时间变量。
📚 2. BISHENG 平台知识库管理和体验优化
<img width="1280" height="672" alt="image" src="https://github.com/user-attachments/assets/a0a68eb9-0cc1-42eb-9c17-4cfaa2bd45c7" />
2.1 ✨ UI/交互全面优化: 信息更清晰、操作更流畅,支持批量操作与已上传文件的分段策略调整。
2.2 🔄 文档知识库支持切换 Embedding 模型并重建。
🧩 3. 工作流功能优化
3.1 🖼️ 报告节点增强: Word 模板支持插入 Markdown 图片/图片文件、Markdown 表格/表格文件。
3.2 👤 开始节点新增 userinfo:返回用户名、用户组、角色信息**,便于做基于用户身份的逻辑处理。
3.3 🗄️ 助手节点数据库查询扩充: 新增对 DB2、PostgreSQL、GaussDB、Oracle 的查询支持。
3.4 🏷️ 工作流 API 回参增强: 新增节点名称字段 nodename**,便于定制化处理与日志追踪。
3.5 🔧 工具节点入参更智能:支持自动数据类型转换,降低参数校验与清洗成本。
🤖 4. 模型功能优化
4.1 🧪 微调能力升级:LLaMA-Factory 升级至最新稳定版本;可微调模型列表同步扩展至当前主流新模型。
4.2 🔌 模型管理拓展: 新增 华为 MindIE 接口格式。
🛠️ 5. 平台其他能力优化
5.1 🧠 本地推理模型效果渲染:Ollama、MindIE 等本地推理框架思考内容支持 <think> 样式渲染。
5.2 🛰️ 助手 API 实现实时流式响应:显著优化等待体验与交互反馈。
5.3 🔗 工具生态更开放: 支持streamable 协议 MCP 工具接入。
5.4 🧪 评测能力增强: 评测功能支持一问一答类型的工作流。
🐞 6. 优化会话头像和应用头像不一致、Milvus 操作结果未生效等若干缺陷和体验。