​
How it Works

When you add the tabular analysis tool and the specified context to your API request, Zylon follows these steps:
  1. Analyzes query intent: Zylon understands your analytical question and determines what specific data analysis needs to be performed.
  2. Processes tabular data: Executes analysis using the tool against the ingested tabular data, focusing on numerical calculations and data insights.
  3. Returns analytical results: Provides direct analytical answers based on the data processing and calculations performed.

​
Important Notes

  • To use this tool, you must specify the type tabular_analysis_v1.
  • The tool_context must contain at least one artifact of type ingested_artifact.
  • Artifacts must contain tabular data such as CSV files, Excel spreadsheets, PDFs with tables, or other structured data formats.

​
Example Usage

In this practical example, we upload a CSV file containing stock price data and then perform analytical queries to find specific insights.
1

Upload the Document

First, you need to upload a CSV file or similar tabular document that contains the data you want to analyze. For detailed instructions on how to ingest documents, refer to the Ingest Content API documentation.
2

Perform Tabular Analysis

Once the document is processed, we perform the analytical query by specifying messages, tools, context, and tool choice in the request body.
curl --location 'https://<your-host>/api/gpt/v1/messages' \
--header 'Authorization: Bearer <your-token>' \
--header 'x-org: default' \
--header 'Content-Type: application/json' \
--data '{
  "messages": [
    {
      "role": "user",
      "content": "What is the day with the highest closing price?"
    }
  ],
  "tools": [
    {
      "name": "tabular_analysis",
      "type": "tabular_analysis_v1"
    }
  ],
  "tool_choice": { "type": "auto" },
  "tool_context": [
    {
      "type": "ingested_artifact",
      "context_filter": {
        "collection": "stock_documents",
        "artifacts": [
          "8853b8c6-59ea-4874-b0a2-fe96cc15eb13"
        ]
      }
    }
  ]
}'
3

Response

The system responds with the analytical processing performed and the direct answer to your query:
 {
   "id": "msg_978e1f57031a41b2bc9140dfe2696ccd",
   "type": "message",
   "role": "assistant",
   "content": [
       {
           "type": "tool_use",
           "start_timestamp": "2025-09-24T16:38:02.179513Z",
           "stop_timestamp": "2025-09-24T16:38:03.487913Z",
           "id": "toolu_01997c96cec37ccc9aafac2566b8c60f",
           "name": "tabular_analysis",
           "input": {
               "query": "Find the day with the highest closing price"
           }
       },
       {
           "type": "tool_result",
           "start_timestamp": "2025-09-24T16:38:16.028423Z",
           "stop_timestamp": "2025-09-24T16:38:16.030538Z",
           "tool_use_id": "toolu_01997c96cec37ccc9aafac2566b8c60f",
           "content": [
               {
                   "type": "text",
                   "text": "The day with the highest closing price is 2024-12-26."
               }
           ],
           "is_error": false
       },
       {
           "type": "text",
           "start_timestamp": "2025-09-24T16:38:16.341826Z",
           "stop_timestamp": "2025-09-24T16:38:17.743287Z",
           "text": "The day with the highest closing price is **2024-12-26**."
       }
   ],
   "model": "private-gpt",
   "stop_reason": "end_turn",
   "stop_sequence": null,
   "usage": {
       "input_tokens": 621,
       "output_tokens": 22
   }
}