From Bright Pattern Documentation

Revision as of 20:46, 9 November 2026 by Laura.donovan (talk | contribs) (Updated via BpClonePage extension. Source page: draft:AI-Agent-2.0/ExecutionHistory)

< Previous | Next >
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Execution History

Click on the Execution History sidebar link to review AI Flow performance in testing and production.

At the top of the page, select All Types, Production, or Debug to choose which executions to display.

You can view a particular AI Flow run's execution history by entering the run or external ID in the Execution History search bar.

Column Name Sort By? Description
Flagged yes Click the flag icon to flag this run for attention
Run ID no Unique identifier associated with this execution internally
External ID no Unique identifier associated with the system that triggered this execution (e.g. BPCC interaction record). For internal runs, this field reads Debug instead.
Status yes
  • Abandoned by customer
  • Completed
  • Failed due to unrecoverable error or connectivity problem
  • Objective judged Incomplete in the allotted time by AI Agent lacking context, capabilities, or instructions
  • Running
Date yes Date and time the interaction began
Initiated By yes Who initiated the interaction:
  • Trigger name, for AI Flows invoked by external systems
  • AI Agent 2.0 username, for internal test runs
AI Flow yes AI Flow associated with this execution
Execution Details no AI-generated interaction summary
Duration yes Total execution time

Filters

Click the Filters button to filter execution history by one or more of the following:

  • Certain table properties, including status, date range, initiating party, AI Flow, flagged status, etc.
  • The channel through which they were invoked: Webhook, WebSocket, or Test AI Flow.
  • Whether hallucinations were detected, and if so, the type of hallucination; AI Evaluator type; or executions with a specific AI Evaluator
  • Minimum number of conversation turns

Reviewing Interactions

To review an interaction, click on the relevant Run ID link in the Execution History table.

The top of the page lists interaction metadata such as total token count, input and output values, and highlights from the corresponding Execution History table entry.

Chat

The Chat section displays all conversation turns exchanged in the interaction.

Steps

The Steps section lists each agent, tool, and evaluator called during the interaction. Each step displays the corresponding timestamp, execution time, and token count. Click on a step to view the corresponding input and output data.

For AI steps, hover your cursor over the icon to the left of the timestamp to view the AI model used.

Evaluator steps indicate whether an AI response was approved or denied. Click on an evaluation step showing an X icon to examine the evaluation feedback given.

Input

JSON input data given to the AI Agent or tool invoked during the selected step. Enable the Verbose toggle to view all technical details.

Output

JSON output data returned by the AI Agent or tool invoked during the selected step. Enable the Verbose toggle to view all technical details.

< Previous | Next >