Overview
The Prompt Execution History Dashboard provides visibility into all past executions of prompts that were run through the platform using any supported AI model. This dashboard helps users understand how their prompts behave in real execution scenarios by displaying detailed information such as inputs, outputs, execution duration, and success status.
Whenever a flow runs a prompt through any Chat Completion step, whether from Anthropic, Azure Foundry, Google Gemini, OpenAI, or the AI Common module, the system automatically logs the full execution details. These records allow users to evaluate prompt behavior, identify unexpected responses, and improve overall prompt design.
Prerequisites
For the Execution History folder to appear under [Project] > Manage > Integrations > AI, the project must have the AI.Common module installed and added as a dependency. Once this module is included, the dashboard becomes available and records prompt execution activity from any Chat Completion step in the system.
Use Case
Users who design or maintain prompts often need insight into how those prompts perform over time. If a prompt begins returning unexpected, inconsistent, or low-quality outputs, the dashboard provides a way to review exactly what occurred during each execution - what input was used, which model processed it, how long it took, and what response the model returned.
By reviewing this history, users can identify patterns, troubleshoot incorrect behavior, and refine prompts to improve accuracy, quality, and consistency.
Execution History Details
The dashboard includes the following columns for each recorded execution:
- Prompt – The name or identifier of the prompt used during execution.
- Model – The AI model used by the Chat Completion step (Anthropic, Azure Foundry, Google Gemini, OpenAI, or open models via AI Common).
- Execution Time (ms) – The total time taken by the model to process the prompt.
- Successful – Indicates whether the execution completed successfully (True) or encountered an error (False).
- Content – The actual prompt text provided to the model.
- Response – The output returned by the model based on the prompt.
- Ran At – The date and time of the execution.
- Provider Name – The AI provider used for the execution.
- Feature Type – Indicates whether the execution was a Chat Completion or an Agent-based execution.
- Agent ID – The unique identifier of the Agent (if applicable).
- Agent Name – The name of the Agent involved in the execution.
- Agent Loop Count – The number of iterations performed during the Agent execution.
These execution logs provide full context around each prompt and agent execution, which is essential for debugging, analysis, and long-term optimization.
Data Source Information
The data displayed in the Prompt Execution History Dashboard is sourced from the dbo.aiprompt_execution_history table. This table stores all recorded prompt execution activity captured from Chat Completion and Agent-based steps across supported AI modules.
The table includes additional fields for agent executions, such as ProviderName, FeatureType, AgentId, AgentName, and Agent_loop_count. These fields are stored at the database level and can be included by editing reports.
Chat Completion executions are no longer recorded in the AI_Audit table and are instead captured in the execution history table.
Users may also create their own custom reports, dashboards, or data views by querying this table directly, allowing deeper analysis and expanded insight into AI prompt performance.
