Test data and conversation context for Testing Center
Key Takeaway
AI-generated test cases don't use conversation history or context variables. For tests that need earlier turns or runtime context, select "Include conversation history" and "Include context variables", fill in the added columns in the CSV template, and upload the file.
Testing Center runs an agent against a batch of test cases to check that it picks the right subagents and actions and gives accurate answers. You can have AI generate test cases from the agent's subagents and actions or from its knowledge. You can also upload your own cases in a CSV file. Uploaded CSV tests can carry conversation history and context variables, so each test behaves more like a real conversation.
Three ways to supply test data
Testing Center comes in two versions: the legacy Agentforce Testing Center in Setup and the new Testing Center in Agentforce Studio (Beta). Both versions give you the same three ways to supply test data.
| Use it to | How it works | |
|---|---|---|
| Generate tests from subagents and actions | Test specific intents or workflows, especially after updates or before launch | Finds the subagents and actions the agent can use, drafts utterances from their instructions to trigger them, then checks that the agent selected the intended subagent and action |
| Generate tests from knowledge sources | Confirm accurate, content-driven answers from connected sources such as articles or PDFs, for example company policies or product specifications | Creates question-and-answer style test cases; needs an agent connected to an Agentforce Data Library |
| Upload a CSV file | Get full control, such as for custom test scenarios or validating specific business logic | You download the CSV template, fill it in with your scenarios, and upload it |
Volume limits differ by version:
- In the legacy Testing Center, AI generation creates up to 100 test cases at a time.
- In the legacy Testing Center, an uploaded CSV can run up to 1000 tests per file.
- In the new Testing Center (Beta), knowledge-based generation creates up to 200 test cases at a time.