Release Notes
2026.05.26
Version Updates
- chatbi web: 4.8.7
- chatbi server: 2.13.1
- chatbi datamind: 2.20.0
Feature Updates
Frontend Features
| Feature | Summary | Related Document |
|---|---|---|
| Multilingual Display for L1 Analysis Parameters | When the system language is English and the user asks questions in English, analysis parameters can now be displayed in English. | Show / Hide Analysis Parameters |
Management Features
| Feature | Summary | Related Document |
|---|---|---|
| Enhanced Theme Avatar Configuration | When configuring a theme avatar, you can now generate one randomly and quickly select an icon and background color. | Agent Display Settings |
| Enhanced Testing Features | AI can now help generate test questions by combining dataset fields, custom generation preferences, metric definitions, and similar inputs within the current theme. | Test Q&A Effect |
| Enhanced Q&A Sample Rewriting | Q&A Sample rewriting now supports recalling both general knowledge and business knowledge for more accurate rewriting results. | Sample Set Rewrite |
| Trace Export | Domain-level administrators can export question records and processing traces within the current domain. This feature is available only in private deployment. | Trace Export |
2026.05.19
Version Updates
- chatbi web: 4.8.6
- chatbi server: 2.13.0
- chatbi datamind: 2.19.0
Feature Updates
Management Features
| Feature | Summary | Related Document |
|---|---|---|
| Multiple Query Usage Rules | Different query usage rules can now be configured for multiple user groups to enable finer-grained usage control. | Management Center |
| Automatic Numeric Scaling | Numeric values can now be scaled automatically based on their length, for example 12304 can be displayed as 12.304K. If scaling has already been handled in the dataset, do not enable this feature. | Basic Information |
2026.05.12
Version Updates
- chatbi web: 4.8.5
- chatbi server: 2.12.1
- chatbi datamind: 2.18.0
Feature Updates
Frontend Features
| Feature | Summary | Related Document |
|---|---|---|
| Voice Q&A on PC | After enabling Voice QA Switch Configuration at the domain level, users can ask questions by voice on PC. | Query Agent |
| Input Box Interaction Optimization | The PC input area has been optimized. Users can now quickly view Common Questions and Detailed Information, including dataset field information for the current theme. | Query Agent |
| File Upload for L2 Analysis | In L2 insight analysis, users can upload Excel and CSV files for data analysis. | Insight Agent |
Management Features
| Feature | Summary | Related Document |
|---|---|---|
| Separate Opening Recommendations for L1 and L2 | Different opening recommendation questions can now be configured separately for L1 Query and L2 Insight. Both Fixed Questions and Model-generated Questions are supported. | Basic Information |
| Guess You May Ask Recommendation Logic | Administrators can now define the recommendation logic for Guess You May Ask. After a user asks a question, the system combines the user input with the configured strategy to generate additional questions that may interest the user. | Basic Information |
| Table Relation Knowledge | A new Table Relation feature allows administrators to maintain relationships across multiple data tables through drag-and-drop configuration, providing knowledge support for multi-table analysis. | Table Relation |
2026.04.28
Version Updates
- chatbi web: 4.8.4
- chatbi server: 2.12.0
- chatbi datamind: 2.17.0
Feature Updates
Frontend Features
| Feature | Summary | Related Document |
|---|---|---|
| New Feedback Entry in the Frontend | Users can now review the content they previously downvoted and the feedback handled by administrators. | Query Agent |
| L2 Data Volume Limit Alert | The prompt shown when L2 data volume exceeds the system limit, meaning more than 1000 cells are interpreted, has been optimized to reduce user anxiety. | Insight Agent |
Management Features
| Feature | Summary | Related Document |
|---|---|---|
| Planning Mode | Insight Agent now supports Planning Mode. In this mode, the agent confirms the analysis approach with the user before generating a report. | Insight Agent |
| BI Dataset Exception Display | Datasets that still contain issues after BI dataset quality inspection are now highlighted in the connected dataset list. | Basic Information |
| Email Push Switch Control | A new domain-level switch controls Email Push. When disabled, email channels are hidden in data Q&A, insight reports, and subscription management. | Management Center |
| Voice Model Selection | The voice model service used for voice questions can now be configured on the page. | Management Center |
2026.04
Feature Updates
| Feature | Summary | Related Document |
|---|---|---|
| Fast Mode Visualization | Data can now be visualized when answering questions in Fast mode. | Management Center |
| Highlight Numeric Differences | After Query Agent returns a chart, users can choose whether the axis should start from zero. | Query Agent |
| Image-based Questions | Query Agent now supports uploading images to ask questions. | Query Agent |
| Data Confidence Display | Query Agent can now display analysis parameters together with the data result, including filters, dimensions, and metric names from the SQL, making the result easier to trust and understand. | Query Agent |
| Common Phrases Optimization | Users can now copy a question with one click, add a question to Common Phrases with one click, and type / in the input box to insert saved phrases quickly. | Query Agent |
| High-precision Mode | A new High-precision mode is available. If selected by the theme administrator, multiple SQL statements are generated for each question. The system then cross-validates them through the model and executes the SQL with the highest confidence score to improve answer accuracy. | Basic Information |
| Trace Tracking | In the dialog record management list, Langfuse trace records can now be viewed for easier troubleshooting. | Usage Tracking |
| Agent Import and Export | Agents can now be imported and exported, making migration easier in private deployment environments. | Create and Configure an Agent |
| Agent Testing | Testing has been enhanced with support for single-item testing, previewing more test data, and improved export capabilities. | Create and Configure an Agent |
| Database Access | Support has been added for direct Kylin datasets. | / |
| Enhanced Dialog Records | Dialog records now include a Process Exception indicator to help administrators identify issues faster and improve operational efficiency. | Usage Tracking |
| Dataset Management | When adding datasets to an agent, users can now choose which fields to disable. | Basic Information |
| Q&A Sample Rewriting | A new SQL rewriting feature based on Q&A Sample is available. After a frontend user submits a question, the system retrieves and matches relevant examples from the sample set. If the match is strong, the system rewrites SQL based on the correct SQL from the sample, improving answer accuracy. | Query Agent |
| Export Dialog Records | The dialog record list now supports export, including Q&A status and failure reasons. | Usage Tracking |
| Analysis Specialist | Insight Agent now supports Analysis Specialist, allowing teams to manage and accumulate analysis methods for specific business domains. Frontline business users can mention a specialist to generate reports with more stable insights and better compliance with domain knowledge. | Insight Agent |
| Hide Email Entry | Administrators can now customize whether the email sending entry is shown or hidden. | Management Center |
| SQL Copy | In the L2 data query tool, users can now view SQL details and copy them with one click. | Insight Agent |
Product Behavior Changes
UI Page Restructure
The entry points have been consolidated. The Query page, Insight page, and Agent Management page are now merged under Data Agent.
Frontend Entry Changes
Before the upgrade: Nine-grid Menu > chatbi Frontend
After the upgrade:
-
Query:
Data Agent > Query Agent
-
Insight:
Data Agent > Insight Agent
2025.10
Feature Updates
| Feature | Summary | Related Document |
|---|---|---|
| Manage Like/Dislike Records in Dialog History | Administrators can now like or dislike records, submit feedback, and review user feedback records. | Usage Tracking |
New Fast Mode | When the analysis mode is Query Analysis, the frontend input box now includes a Fast mode entry that lets the model provide answers more quickly. | Query Agent |
| Pivoting a Result Table by a User-selected Field | When users switch a query result to table view and the data contains at least two dimensions, one dimension can now be converted into a comparison field in the frontend. | Usage Tracking |
| Configurable Retention Days for Historical Query Results | A Query Data Retention Days option has been added to configuration, with a default of 30 days. Query results older than that are cleaned from dialog history. | Management Center |
| Insight Report Interaction Optimization | After generation, reports are collapsed by default and can be opened on the right side. Dialog-related actions such as Favorite, Like/Dislike, and Copy Message ID remain in the conversation area, while report-related actions such as Export and Email appear in the report area. | Insight Agent |
| SQL Limit Exceeded Alert | Query Analysis Row Limit is now 1000 rows by default and can be expanded up to 100,000 rows. | Management Center |
/chatBI Opens the Most Recently Visited Agent by Default | Visiting the /chatBI route now opens the agent Q&A screen most recently used by the current user on both PC and mobile. | Query Agent |
Product Behavior Changes
Insight Report Entry Changes
Before the upgrade: the report was displayed directly, and all feature buttons appeared below the report.
After the upgrade: the report is displayed on the right, and the Export and Email buttons have been moved to the upper-right corner of the report.
Landscape Visualization Support on Mobile
Before the upgrade: after the user unlocked the screen and rotated the device, the table could not be displayed horizontally.
After the upgrade: after the user unlocked the screen and rotated the device, the table is displayed horizontally automatically.
/chatBI Opens the Most Recently Visited Agent by Default
Before the upgrade: every visit to the /chatBI route opened the first agent Q&A screen that the user had permission to access.
After the upgrade: visiting the /chatBI route opens the agent Q&A screen most recently used by the current user on both PC and mobile.
2025.09
Feature Updates
| Feature | Summary | Related Document |
|---|---|---|
| Exposure of Troubleshooting Chain Information for Data Retrieval | During query analysis, users can now see the table knowledge, business knowledge, and Q&A Sample content sent to the model. | Usage Tracking |
| Custom Recommended Questions | Recommended questions can now be customized, with up to 20 examples. The model uses these example questions together with the current user's dialog history and the agent's data structure to generate 3 sample questions. | Basic Information |
| Knowledge Recall Count | The number of times a business knowledge item has been recalled is now displayed, making it easier to identify frequently used knowledge. | Knowledge Management |
| User Attribute Awareness in the Q&A Flow | When a user asks a question, the user's attribute values are passed to the model so it can generate SQL more accurately. | Basic Information |
Product Behavior Changes
Operation Log Removed from Dialog History and Replaced by Retrieved Knowledge
Before the upgrade:
In the query history dialog, the upper-right corner displayed Operation Log, Notification History, and Add to Q&A Sample.
After the upgrade:
In the query history dialog, the upper-right corner displays Retrieved Knowledge, Notification History, and Add to Q&A Sample.
Custom Recommended Questions for the Model
Before the upgrade:
The model generated recommended questions only from dataset descriptions and dataset field information.
After the upgrade:
The model now generates 3 sample questions based on recommended question examples, the user's current dialog history, and the agent's internal data structure. Up to 20 sample questions can be configured.
Automatic Knowledge Generation by the Model
Before the upgrade:
Knowledge marked as incorrect during testing had to be corrected manually.
After the upgrade:
During testing, clicking Generate Knowledge starts automatic knowledge generation and validation for items currently marked as incorrect in Review Result, and the progress is displayed during the process.
2025.08
Feature Updates
| Feature | Summary | Related Document |
|---|---|---|
Added Retrieved Knowledge and Add to Q&A Sample to the Attribution Flow | In Insight Analysis Process, clicking Data Query now provides Add to Q&A Sample and Retrieved Knowledge. | Usage Tracking |
| Exposed Query Thinking Process | During query analysis, after relevant table information is displayed, the system now shows the thinking process below it while still rendering charts normally, helping users understand the data more intuitively. | Query Agent |
| SQL Explanation in Query | SQL execution logic, filter conditions, and data relationships can now be broken down and explained. | Query Agent |
| Enhanced Dataset Learning and Optimized Field Enum Logic | Dataset learning can now be triggered manually, and the logic for field enumeration values has been adjusted. | Basic Information |
| Enable/Disable Knowledge Support | Q&A Sample and Business Knowledge now support both individual and batch enable/disable management. | Knowledge Management |
| Configurable LLM Service | Administrators can now configure and test the LLM service in the Data Agent management backend to support private deployments that connect to an internal LLM. | Management Center |
Product Behavior Changes
Exposed Thinking Process in Query (thoughts)
Before the upgrade: the thinking process was not displayed while the answer was being generated.
After the upgrade: while the answer is being generated, the thinking process is displayed below the related table information so users can understand how the system reasons before querying, after first understanding the table structure and field meanings.
SQL Explanation Support in Query
Before the upgrade: clicking SQL only copied the current SQL.
After the upgrade: clicking SQL not only copies the current SQL, but also allows users to view SQL Explanation in the SQL Preview dialog.
Enhanced Dataset Learning and Optimized Field Enum Logic
Before the upgrade: dataset learning could only be triggered automatically every 30 minutes.
After the upgrade: dataset learning can now be triggered both automatically and manually. The automatic trigger frequency has been changed to once per day.
UI Visual Adjustments: Recall Style, Dialog Input, Tool Calls, and AI Usage Reminder
| No. | Update Description |
|---|---|
| 1 | Agent selection is now collapsed into a separate page. The left-side double-column layout, consisting of agent selection and dialog history, has been changed to a single-column layout for dialog history only. |
| 2 | Insight tool call labels have been updated to Data Query, Knowledge Retrieval, Code Execution, and Web Search. |
| 3 | The dialog input now supports up to 1000 characters. |
| 4 | A gray reminder has been added below the dialog box: You are using an AI tool. Please protect personal privacy and sensitive data, and carefully verify AI-generated content. |
| 5 | The New Session entry has been moved below the dialog. If the user has already entered content and clicks New Session, the content is carried into the new session input box, and the Insight Analysis or Query Analysis selection is preserved. When switching between these two analysis modes in the first dialog, the behavior is the same. |
| 6 | A quick-scroll entry has been added when the dialog is scrolled upward. Clicking it jumps directly to the bottom. |
| 7 | A Stop Generating entry has been added during the session so model output can be interrupted. |
5. Entry for Generating Business Knowledge Removed
Before the upgrade: when creating a new agent or using the business knowledge library, knowledge could be generated in batch by selecting dashboards.
After the upgrade: this entry has been removed.
2025.07
Feature Updates
| Feature | Summary | Related Document |
|---|---|---|
| Learning Status for Q&A Sample and Business Knowledge | The learning status of each knowledge item is now displayed so users can understand learning progress more intuitively. | Knowledge Management |
| Feature Switch Configuration | Users can now decide whether to enable Recommended Questions, Guided Questions, Personalized Memory, and Insight features. | Basic Information |
| Frontend Visual Upgrade | The frontend interaction experience on mobile and PC has been optimized. | Query Agent |
| Dashboard Recommendation in the Backend | After a user asks a question, the system automatically retrieves the 3 chart cards most relevant to the question from the linked dashboards and displays them as recommendations for more intuitive data insights. | Basic Information |
| Personalized Memory in the Frontend | Users can now choose whether to enable Personalized Memory in the frontend. Enabling it means the current user allows Q&A to reference saved personalized memory. | Basic Information |
Product Behavior Changes
Learning Status Display
Before the upgrade: the operation list for Q&A Sample and Business Knowledge displayed only the most recent update time and the most recent operator.
After the upgrade: the operation list now displays learning status, the latest completed learning time, and the most recent operator.
Agent-level Switch Control in Data Agent
Before the upgrade: no switch control was available.
After the upgrade: custom switch control is now supported.
Visualization Optimization for Tables and KPI Cards
Before the upgrade:
In table scenarios, when there was only one row of data, dimensions and values were converted according to simple text-type and numeric-type rules.
After the upgrade:
In table scenarios, dimension recognition is now determined by the model.
The default chart recommendation rules have also been updated:
- when the result contains one row and one column, a
KPI Cardis recommended by default; - when the result contains one row and multiple columns, a
Comparison KPI Cardis recommended by default; - if no chart satisfies the recommendation conditions, a table is recommended by default.
Visualization Agent Optimization - Chart Display
Before the upgrade: axes followed the default ECharts generation rules.
After the upgrade: axis ranges are now generated by the model according to the data range.
Always Open Message Reply Links in a New Tab
Before the upgrade: links in Insight Analysis reply messages were opened in the current tab.
After the upgrade: links in Insight Analysis reply messages are opened in a new tab.
Automatic Rotation for Data Agent Visualization Charts
Before the upgrade: in the new visualization style, axis labels were laid out flat and could be truncated.
After the upgrade: axis labels are rotated automatically in the new visualization style.
Search Attribution Messages by Message ID in Dialog History
Before the upgrade: users had to switch between query and insight types before searching, and insight messages could not be searched by message ID.
After the upgrade: query and insight messages are now displayed together and can be filtered in the filter bar, and insight messages can also be searched by message ID.
2025.06
Feature Updates
| Feature | Summary | Related Document |
|---|---|---|
| Insight Analysis Released | A new Insight Analysis button has been added to the frontend. After selecting it, users can perform deeper analysis of business questions such as How has recent sales performance been?, providing stronger support for decision-making. | Insight Agent |
Multilingual Support for Data Agent | Data Agent now supports multiple languages. After multilingual support is enabled and the language is switched on the BI side, the Data Agent page switches accordingly. | BI Platform Preparation Before Configuring Data Agent |
| Cascade Updates for Q&A Sample SQL and Test-set SQL When Datasets Change | When dataset names or field names change and dataset table knowledge is updated, related Q&A Sample SQL and test-answer SQL are updated automatically, without requiring users to modify them manually. | Basic Information |
| Exposed SQL Error Information | During Test > Review, users can now view and copy the message ID, review the operation log, and copy the executed SQL. | Create and Configure an Agent |
| Richer ChatBI Welcome Message Styling | ChatBI welcome messages now support rich text editing and references to specific attributes. | Basic Information |
| Knowledge Tag Management | Q&A Sample and Business Knowledge now support tag management, making it easier to organize and search content in the knowledge base. | Knowledge Management |
Product Behavior Changes
Cascade Updates for Q&A Sample SQL and Test-set SQL When Datasets Change
Before the upgrade:
When dataset names or field names changed, related Q&A Sample SQL and test-answer SQL that referenced the original names were not updated, and they could still be recalled during actual Q&A, which negatively affected answer accuracy.
After the upgrade:
When dataset names or field names change, once dataset table knowledge is updated, related Q&A Sample SQL and test-answer SQL are updated automatically and no manual adjustment is required.