Skip to main content

Execution History

Track how your prompt templates are being used in production.

Overview​

Execution History shows the last 10 calls from production with full details:

  • Token usage per call
  • Response times
  • Model and version used
  • User who triggered it
  • Full request/response data

Execution History View​

What's Tracked​

Per Execution​

FieldDescription
TimestampWhen template was executed
VersionWhich template version
ModelAI model used
TokensInput + output tokens
UserWho triggered execution
VariablesInput values provided
ResponseAI-generated output
DurationResponse time

Use Cases​

1. Debugging Production Issues​

Scenario: Agent is getting wrong responses

Steps:

  1. Open template in GenAI Explorer
  2. Scroll to Execution History
  3. Find the problematic execution
  4. Click "View Details"
  5. Check variable values
  6. Identify the issue (empty, wrong, malformed)
  7. Fix and test

2. Monitoring Token Usage​

Scenario: Want to reduce costs

Steps:

  1. Review last 10 executions
  2. Note average token count
  3. Identify high-token calls
  4. Look for patterns (long variables, verbose output)
  5. Optimize template

See: Token Optimization for strategies

3. Version Performance Comparison​

Scenario: New template version deployed

Steps:

  1. Compare token counts between versions
  2. Check response times
  3. Review output quality
  4. Decide if new version is better

Execution Details​

Click "View Details" to see:

{
  "executionTime": "2025-12-10T10:30:00Z",
  "templateVersion": "v2",
  "model": "GPT-4 Omni",
  "tokens": {
    "input": 150,
    "output": 300,
    "total": 450
  },
  "duration": "2.3s",
  "user": "john.smith@company.com",
  "variables": {
    "Customer_Name": "John Smith",
    "Product": "Enterprise License"
  },
  "finalPrompt": "You are a customer service agent...",
  "response": "Hello John Smith! I'd be happy..."
}

Full History in Data Cloud​

The last 10 executions are shown in GenAI Explorer.
For complete history, query Data Cloud:

Example Query
SELECT 
  req.gatewayRequestId__c,
  req.prompt__c,
  req.maskedPrompt__c,
  req.userId__c,
  req.totalTokens__c,
  req.promptTokens__c,
  req.completionTokens__c,
  gen.responseText__c,
  req.promptTemplateDevName__c,
  req.promptTemplateVersionNo__c,
  req.timestamp__c,
  req.model__c,
  req.provider__c
FROM
  GenAIGatewayRequest__dlm req,
  GenAIGatewayResponse__dlm res,
  GenAIGeneration__dlm gen
WHERE 
  req.gatewayRequestId__c = res.generationRequestId__c
  AND res.generationResponseId__c = gen.generationResponseId__c
  AND req.promptTemplateDevName__c = '"${id}"'
ORDER BY req.timestamp__c DESC
LIMIT 10

See: Data Cloud Integration for more query examples

Regular monitoring ensures templates perform well and stay cost-effective.