Read saved project files with Task analysis
Review a task plan, approve project access and data transfers, then inspect the answers and run evidence.
Use a Cloud organization with AI access.
Use Task analysis in a Cloud 2.1.2.10 project to analyze saved files with your own OpenAI API key. Save the files you want to review before starting.
Connect your OpenAI model
In the project’s activity rail, choose AI. In the chat header, choose AI settings (the gear icon), then choose Connections.
In the OpenAI row, choose Connect (or Replace key for an existing connection), enter your OpenAI API key and choose Save key. Confirm that Key saved appears after saving.
Return to the chat, type
/, and choose Task analysis.
Review the task plan
Enter an Overall goal. For the first task, enter a Task title and an Analysis request.
In Allowed saved files (one path per line), enter each project-relative file path on its own line. Enter one criterion per line in Review criteria (one per line).
To analyze another task, choose Add task. Use Wait for tasks to select any tasks that should finish first.
In Model, choose an available supported OpenAI Responses model from your connection. Set Reasoning effort to
low,mediumorhigh. Review Concurrent tasks, Total model requests, Total file or diagnostic reads and Time limit (seconds).Review every task and choose Review and start analysis.
Approve access and review the evidence
At the project-access request, open View details and review the model, permitted reads and task request. Choose Allow project access to continue.
For each Send tool result request, open View details and inspect the file or diagnostic data prepared for the model. Choose Approve once to send that result, or Deny to decline.
In Analysis results, inspect each task’s status and answer against your review criteria. Open Run evidence to inspect its run identifier and answer hash.
Choose Download run evidence to save the run evidence as JSON. To stop an active analysis, choose Cancel all tasks.