Read an open answer with AI processing
An AI-processed variable reads the free-text answer of a participant. The next page shows the result.
-
Create an AI-processed variable
In Variables, add a variable of the kind AI-processed (
llm-processed). Set these items:- Sources: the data that the model reads. Here, the answer to the input segment,
segment-submission:seg_view:value. The:valueending gives the plain text, not the stored object. - The provider and the model.
- The system prompt: the task, and the exact form of the reply.
- The user prompt: put the source in it with
{{…}}. - Output format: Text keeps the reply as it is. JSON parses the reply.
- Update trigger: Manual. A Process segment then starts the variable.
view_summaryreads the open answer and asks for a stance and a reason of one line. - Sources: the data that the model reads. Here, the answer to the input segment,
-
Run it with a Process segment
Put a Process (AI) segment between the question and the page that uses the result. Select the variable in it.
The participant sees the message of the segment while the model runs. When Carrier stores the result, the study continues. A timeout and a fallback value apply if the call is slow or fails.
The Reading your answer segment runs view_summary. -
Show the result
A subsequent page can use the variable like all other variables. Here, an instruction uses
{{var.view_summary}}. Other items that read variables can also use it, for example an agent prompt.Carrier fills in a page when the page starts. Thus the Process segment must come first, so that the value exists when the page starts.
{{var.view_summary}}goes on the What we heard page, and the study is deployed. -
What the participant sees
The participant writes an answer in their own words. They wait a short time on the Process segment. Then they read how the model understood their answer.
A mixed answer about car-free town centres comes back as Mixed, with a reason of one line.