Lab Experiments
Overviewβ
The Lab Experiments module provides users with a one-stop research assistant capability that automatically executes experimental verification starting from literature or research drafts. Users can create experiment tasks based on uploaded literature attachments, or automatically from channels such as Literature Management and Research Assistant. The system invokes AI agents to complete the full process of literature interpretation, citation retrieval, dataset matching, experimental verification, and result output, achieving a complete closed loop of "Upload Literature β AI Automatic Experiment β Product Output."
The module includes core features such as experiment task management, task execution and progress tracking, experiment product browsing (complete products, output documents, output results), dataset viewing, and related literature preview. It supports two execution strategies β Fast Mode and Expert Mode β and provides a real-time conversation panel on the right side of the detail page, allowing users to continue asking follow-up questions after task execution is successful.
Applicable Rolesβ
All logged-in users. Regular members can only view and operate their own experiment tasks; administrators can view all experiment tasks.
Operation Guideβ
Module Entryβ
Entry: Click "Lab Experiments" in the left sidebar menu to go directly to the page.
The page header displays the module title and description, and below it shows all experiment tasks of the current user in a card grid layout.
View Experiment Tasksβ
After entering the Lab Experiments module, the page automatically loads and displays all experiment tasks in card format. Each card shows the task icon, name, description, update time, and status label.

Operations:
- Search Tasks: Enter task name keywords in the top search box and press Enter to search
- Status Filter: Select a status (Pending, Running, Success, Failed, Stopped) from the dropdown on the right side of the search box to filter
- Refresh List: Click the "Refresh" button to reload the task list
- New Experiment: Click the "New Experiment" button in the top right corner to open the creation dialog
- Load More: Scroll to the bottom of the page to automatically load more tasks; you can also click the expand button at the bottom to load manually
- View Details: Click any card to enter the experiment detail page
New Experimentβ
Steps:
- Click the "New Experiment" button to open the creation dialog
- Click the icon area on the left to select an icon and background color from the popup panel
- Enter the task name
- Enter the task description
- Click or drag to upload an attachment (supports PDF/Word/Markdown formats, max 500MB)
- Click "Create" to submit

Quick Create Experimentβ
In addition to manual creation, experiment tasks can also be automatically created with one click from the following channels. The system automatically fills in the task name and description and associates the corresponding attachments. After automatic creation, the page automatically navigates to the new task's detail page.
Create from Literature List Right-Click Menuβ
- Enter the "Literature Management" module, right-click the target literature in the list
- Select "Create Experiment" from the right-click menu

Create from Literature Detail Pageβ
- Enter the "Literature Management" module, click the target literature to enter the detail page
- Click the "More" operation button at the top of the detail page, select "Create Experiment"

Create from Literature Interpretationβ
- Enter the "Research Assistant" module, select "Literature Interpretation" mode
- Upload or select literature for interpretation
- Click "Create Experiment" on the interpretation result page

Create from Research Explorationβ
- Enter the "Research Assistant" module, select "Research Exploration" mode
- Enter a research question to generate an exploration plan
- Click "Create Experiment" on the plan result page

Edit Experimentβ
On the experiment detail page, click the edit icon next to the title at the top to modify the task name, description, or icon. Click "Save" to apply changes.
Delete Experimentβ
Hover over the target card on the list page, click the delete icon in the top right corner, and confirm to delete.
Note: Only the task creator can delete their own experiments.
Experiment Detailβ
After entering the detail page, a full-screen popup displays the complete information of the experiment task. The left navigation switches between six panels (Research Objective, Related Documents, Datasets, Complete Products, Output Documents, Output Results), and the right side shows the task execution progress panel.
Research Objectiveβ
Displays the original research objective document of the task, helping users confirm the research direction and scope of the experiment.
The Research Objective panel is shown by default after entering the detail page, previewing the task attachment content (i.e., the research objective document).
- The top of the panel shows the source channel information (e.g., "Research Assistant > Uploaded Literature Interpretation", "Literature Management", etc.)

Related Documentsβ
Displays related literature retrieved and cited during the AI experiment process, helping users understand the literature basis of the experiment.
Select "Related Documents" in the left navigation. The left side shows the list of literature associated with the task (title, source, year). Click a literature item to preview the file on the right. Search and filtering within the list are supported.

Datasetsβ
Displays dataset resources matched by AI for the experiment, providing data support for experimental verification.
Select "Datasets" in the left navigation. The page shows the list of datasets matched for the task (name, source platform, description). Datasets with links display an "Open Dataset" button to view in a new window.

Upload Custom Datasetβ
Users can upload local data files as custom datasets to supplement AI-matched data resources.
- Click the "Upload Dataset" button in the dataset panel
- Select a local data file to upload

Complete Productsβ
Displays all product files generated by AI after the experiment is completed, including code, data, configuration files, etc., supporting online preview and download.
Select "Complete Products" in the left navigation. The left side shows the product directory tree (folder/file structure). Click a file node to preview its content on the right (code files support syntax highlighting). Right-click a file node to download a single file, or click the "Download Products" button to package the entire directory tree as a ZIP download.

Output Documentsβ
Displays complete documents automatically extracted by AI from experiment products (such as research reports, analysis summaries, etc.), making it easy for users to read and download directly.
Select "Output Documents" in the left navigation. The left side shows the list of documents extracted from products (PDF/HTML/Markdown formats). Click a document to preview it on the right.

Output Resultsβ
Displays the final analysis results and conclusions of the experiment, serving as the main entry point for users to obtain core experimental findings.
Select "Output Results" in the left navigation. For Markdown format results, the left side shows the title outline and the right side renders the content. Click an outline item to smoothly scroll to the corresponding position.

Execute Experimentβ
After entering the detail page, if the task status is "Pending", click the "Execute Task" button at the top. The right panel will display the AI execution process in real-time streaming. After execution is complete, the product and result data are automatically refreshed.
Task Status Descriptionβ
| Status | Meaning | Available Actions |
|---|---|---|
| Pending | Task created, not yet started | Execute, Edit, Delete |
| Running | AI is executing the experiment | Stop execution (awaiting real-time message output) |
| Success | Execution completed | View products and results, Ask follow-up questions |
| Failed | Execution error | Re-execute, Edit, Delete |
| Stopped | Manually stopped by user | Re-execute, Edit, Delete |
Stop Experimentβ
During task execution, click the "Stop Execution" button at the top of the detail page and confirm to terminate. The task status changes to "Stopped", and you can click "Re-execute" to start again.
Experiment Follow-upβ
When the task execution is successful or the AI sends a clarification message, the input box at the bottom of the right panel becomes available. Enter your follow-up question and click send. The AI returns answers in streaming mode, and you can switch AI models from the dropdown in the bottom right corner.

Note:
- The input box is disabled during task execution; AI Q&A is only available after successful execution.
Research Mapβ
This feature is under continuous iterative development.