Skip to main content
The Response Schema editor helps you set rules for how the LLM should structure its answers. It appears when you select “JSON Schema” from the Response Type dropdown in the Iterate mode. Models Settings Pn JSON Schema is like a blueprint for data. It tells the LLM exactly what information to include and how to format it. This is useful when you need consistent, structured responses. For example, you might want product information always organized the same way with specific details.

​
JSON Schema Configuration Guide

Let’s understand with an example how to set JSON schema response for your prompts.
The Schema that you provide to the model must adhere to the OpenAI JSON schema structure.

​
A reasoning assistant

Response Schema
{
  "name": "simple_response",
  "description": "A simple response with just an answer",
  "strict": true,
  "schema": {
    "type": "object",
    "properties": {
      "answer": {
        "type": "string"
      }
    },
    "required": [
      "answer"
    ],
    "additionalProperties": false
  }
}
System

You are a careful reasoning assistant.

1. Think through the problem in clear, numbered steps.
2. Show every calculation or piece of logic you use.
3. After your reasoning, write “Final Answer: {{your_answer}} on its own line.
4. Do not add anything after that line.
User
Question:
Our SaaS platform can ship only one new feature next quarter. The options are:

Team Analytics – weekly usage reports for managers

AI-Powered Help Bot – answers user questions inside the app

Bulk Import Tool – lets admins upload thousands of records at once

Data:

Engineering capacity: 8 story-points per week for 12 weeks

OKRs: boost Weekly Active Teams (WAT) by +15 % and cut time-to-first-value (TTFV) to < 10 minutes

Current support cost: $12 000 per month; each help ticket ≈ $8

Task:
Using clear, numbered steps, compare impact, effort, and risk for each feature. Decide which single feature we should build to best achieve the OKRs. Finish with “Final Answer: <feature name>” on its own line. Use JSON format
Response
{
  "answer": "1. Establish OKRs and Constraints:\n
    - OKRs: Increase Weekly Active Teams (WAT) by 15% and reduce TTFV to under 10 minutes.\n
    - Engineering capacity: 8 story-points/week x 12 weeks = 96 story-points max this quarter.\n
    - Support cost is high, with each ticket costing $8.\n\n

  2. Compare Features Based on Impact, Effort, and Risk:\n\n

  A. Team Analytics\n
    - Impact: May increase WAT if managers use insights to engage teams, but not guaranteed. Unlikely to reduce TTFV as new users won't see value until reports are generated after use.\n
    - Effort: Medium; requires data aggregation, charting, report scheduling. Approximate effort is 30-40 story-points.\n
    - Risk: Low risk; reporting features are well-known, but less directly tied to OKRs.\n\n

  B. AI-Powered Help Bot\n
    - Impact: Could reduce support tickets by addressing user questions instantly, indirectly reducing support costs. May help reduce TTFV if bot helps onboard users faster.\n
    - Effort: High; requires integration with LLM provider, training, UX work. Estimated at 50-70 story-points.\n
    - Risk: Medium-to-high; LLM accuracy could impact user satisfaction, and training may require ongoing tuning. Some positive impact on OKRs but not guaranteed for WAT.\n\n

  C. Bulk Import Tool\n
    - Impact: Lets admins onboard whole teams faster, directly reducing TTFV (many users get value immediately as data is imported in bulk). Also likely to boost WAT, since more teams get active faster.\n
    - Effort: Medium; file upload, mapping, error handling, validation. Approximate effort is 40-50 story-points.\n
    - Risk: Medium; data import can be tricky (edge cases), but highly impactful if executed well.\n\n

  3. Synthesis and Recommendation:\n
    - Team Analytics: Less direct impact on the stated OKRs.\n
    - AI-Powered Help Bot: Supports TTFV reduction and support cost, but with higher risk/effort, and unclear WAT impact.\n
    - Bulk Import Tool: Strong direct impact on both WAT and TTFV, within available engineering capacity, with moderate risk.\n\n

  Final Answer: Bulk Import Tool"
}