Create tables

Emit a table artifact with table(...). Provide an array of objects; column names are inferred from object keys.

Reference implementation in this GitHub repository.

Architecture

Emit table artifacts in-line so tabular data renders below the answer.

agents.json configuration:

return JSONResponse(content={
  "vanilla_agent_tables": {
    "endpoints": {"query": "http://localhost:7777/v1/query"},
    "features": {
      "widget-dashboard-select": False,
      "widget-dashboard-search": False,
    },
  }
})

Query flow

• Process user request and generate data (from widgets, analysis, or computation)
• Stream explanatory text with message_chunk()
• Create table data as list of dictionaries (column names from object keys)
• Emit table() artifact with data, name, and description
• Table renders below the streamed text in Workspace UI
• Continue with additional content or complete response

OpenBB AI SDK

• table(data, name, description): Creates MessageArtifactSSE for table display
  • data: List of dictionaries where keys become column headers
  • name: Table title displayed in UI
  • description: Optional table description
• message_chunk(text): Streams explanatory text before/after tables
• Tables automatically handle data formatting and sorting in UI

Core logic

from openbb_ai import message_chunk, table
from openbb_ai.models import QueryRequest

async def query(request: QueryRequest) -> EventSourceResponse:
    async def execution_loop():
        # Stream introduction
        yield message_chunk("Let me analyze the financial data and create a summary table.\n\n").model_dump()

        # Generate or process data (from widgets, calculations, etc.)
        financial_data = [
            {"symbol": "AAPL", "price": 150.25, "change": 2.5, "volume": 1200000},
            {"symbol": "MSFT", "price": 280.75, "change": -1.2, "volume": 890000},
            {"symbol": "GOOGL", "price": 2650.80, "change": 15.3, "volume": 560000},
        ]

        # Create table artifact
        yield table(
            data=financial_data,
            name="Stock Market Summary",
            description="Current stock prices with daily changes and trading volume"
        ).model_dump()

        # Stream additional analysis
        yield message_chunk("\n\nThe table above shows the current market status. AAPL and GOOGL are up, while MSFT is down slightly.").model_dump()

    return EventSourceResponse(execution_loop(), media_type="text/event-stream")