Skip to main content

Server-Side Rendered Mode (SSRM)

Server-Side Rendered Mode (SSRM) is an advanced widget implementation approach designed to handle large datasets efficiently in OpenBB Workspace. Unlike traditional widgets that load all data at once, SSRM mode enables server-side pagination, filtering, and sorting, making it ideal for displaying massive datasets without compromising performance.

Key Features & Use Cases​

SSRM mode is designed for enterprise applications that need to handle large datasets (200,000+ rows) efficiently. Instead of loading all data at once, it provides server-side operations that optimize performance and user experience.

How SSRM Differs from Standard Table Widget​

AspectStandard Table WidgetSSRM Mode
Data LoadingAll data loaded at onceData loaded on-demand
FilteringClient-side filteringServer-side filtering
SortingClient-side sortingServer-side sorting
Memory UsageHigh for large datasetsOptimized memory usage
Initial Load TimeSlower for large datasetsFast initial load
Network TrafficHigh initial payloadOptimized payloads
ComplexitySimple implementationAdvanced server logic

Widget Configuration​

To use SSRM mode, your widget configuration must specify the correct widget type:

{
  "name": "SSRM Data Table",
  "description": "Server-side rendered table for large datasets",
  "type": "ssrm_table",
  "endpoint": "data-ssrm",
  "gridData": {"w": 12, "h": 8}
}

Important: The widget type must be set to "ssrm_table" to enable server-side row model functionality. This tells OpenBB Workspace to use the SSRM interface instead of the normal table widget.

Implementation Overview​

The SSRM example implementation is built as a FastAPI server that handles AgGrid's Server-Side Row Model requests.

For complete implementation examples and code samples, refer to the SSRM Mode Examples Repository.

Here's what the implementation provides:

Core Architecture​

  • FastAPI Server: Main application handling SSRM requests from the widget
  • Database Manager: Multi-database interface supporting SQLite, MySql, and Snowflake
  • Custom Models: AgGridOptions and AgRows models for structured data handling
  • Query Engine: Modular system for converting AgGrid requests to optimized SQL queries

Key Endpoint: /data-ssrm​

The main POST endpoint processes AgGrid SSRM requests and handles:

  • Pagination: Processes startRow/endRow parameters for efficient data paging
  • Filtering: Supports text, number, and set filters with various operators (contains, equals, greaterThan, etc.)
  • Sorting: Multi-column sorting with ascending/descending directions
  • Grouping: Row grouping with COUNT aggregation and hierarchical support

Request/Response Flow​

  1. AgGrid frontend sends SSRM request with pagination, filters, sorting, and grouping parameters
  2. Server converts request to a valid query using the AgRows model
  3. Database manager executes query against the database
  4. Results are formatted and returned with rowData and rowCount for AgGrid consumption

AI Data​

When using SSRM mode with AI functionality in OpenBB Workspace, there's an important limitation to understand:

"What You See Is What You Get": AI can only access and analyze the data that is currently visible in the table widget. This means:

  • AI will only see the specific rows that are loaded and displayed on screen
  • If your dataset has 15 million rows but only 100 are currently visible, AI can only work with those 100 rows
  • Pagination, filtering, and sorting affect what data is available to AI analysis
  • AI cannot access the entire dataset that exists in your database

Current Implementation​

This is the current approach for AI data access in SSRM mode. Future implementations are planned to make this more robust and provide broader data access capabilities for AI analysis.

Best Practices for AI Integration​

  • Load relevant data subsets that you want AI to analyze
  • Use filtering to focus on specific data ranges before engaging AI
  • Consider the scope of your analysis when working with paginated data
  • Be aware that AI insights will be limited to the currently displayed dataset

On this page