NLG data

Where does generated data live

Prompt data

When running the first prompt of a workspace, a conversation-set named {workspaceName} - NLG will be automatically created and linked to the workspace being used.
All manual prompt runs (executed from the pinned prompt) will output their data to this conversation set.

warning

Do not unlink the NLG conversation set linked to your workspace. This will prevent you from using prompts.

Pipeline data

When running prompts via pipelines, the generated data lives as part of that pipeline. Each pipeline stores its latest run output. Deleting a pipeline also deletes that pipeline's output.

Exploring generated data

All generated data can be consulted within the Generated tab on the Data view. It is subject to the same filters & controls as your unlabeled data.

Generation run id

Every time a prompt is run the generationRunId of the workspace is incremented. Using the Run filter allows you to filter generated data by run.

Run filter

;

Default to latest run

Via the user preferences, you can set HumanFirst to automatically show the latest prompt run output when navigating to the generated tab.
;

Inspecting generated data

Generated data has slight differences with other data linked to your workspace.

Source conversation

When inspecting a generated utterance, an additional tab (Source conversation) will be present on the inspection panel.
This tab will show the conversation where the utterance used by the prompt came from.

Generated metadata

• ExampleId: the utterance id used to when the prompt was run.
• generationRunId: generation run id the output came from.
• Generation time: when the output was generated, in UTC time.
• integrationId: which integration was used to generate this output.
• llmModelName: which generative model was used by the integration.
• promptId: the prompt used to generate this output.
• sourceConversationId: the conversation id of the utterance used to generate this output.

;