POST
/
api
/
v1
/
synth_data
/
generate
Generate Data From Body
curl --request POST \
  --url https://api.collinear.ai/api/v1/synth_data/generate \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '{
  "model_id": "3c90c3cc-0d44-4b50-8888-8dd25736052a",
  "multiplier": 123,
  "examples": [
    {}
  ],
  "prompt": "<string>"
}'
[
  {
    "original": {},
    "generated": "<string>"
  }
]

​
Alert

This method of generating simulated data is different from the SDK. If you wish to create trait-based simulations, use the SDK

​
Overview

Generate synthetic conversation rows inline by providing example prompts and a few seed examples. The service uses the specified model to fan out new samples that you can immediately download or store. This JSON variant is convenient when building tooling that programmatically defines the schema and instructions.

​
Request body fields

  • model_id: UUID of the model that should author the synthetic responses.
  • multiplier: Integer multiplier applied to the number of examples. If you send three examples and set multiplier to 4, the service targets twelve generations.
  • examples: Array of objects. Each example can include any keys your pipeline understands (commonly conv_prefix, response, context, etc.).
  • prompt: Instruction prompt that explains how the model should transform the examples when creating new rows.

​
Example request

curl https://stage.collinear.ai/api/v1/synth_data/generate \
  -H 'Authorization: Bearer <token>' \
  -H 'Content-Type: application/json' \
  -d '{
    "model_id": "<MODEL_ID>",
    "multiplier": 5,
    "prompt": "Generate safe customer service conversations for billing issues.",
    "examples": [
      {
        "conv_prefix": [
          {"role": "user", "content": "My shipment is missing."}
        ],
        "response": {
          "role": "assistant",
          "content": "I'm sorry to hear that. Let me check the tracking number."
        }
      }
    ]
  }'

​
Response shape

The API returns an array. Each element includes the original example that the model expanded from and a generated string with the newly created row: 
[
  {
    "original": {"conv_prefix": [...], "response": {...}},
    "generated": "{\n  \"conv_prefix\": [...],\n  \"response\": {...}\n}"
  }
]
Parse the generated string into JSON before storing it. The service may stream partial chunks before the final array is delivered; buffer the body before parsing if your HTTP client emits incremental events.

Authorizations

​
Authorization
string
header
required

Bearer authentication header of the form Bearer <token>, where <token> is your auth token.

Body

application/json
​
model_id
string<uuid>
required
​
multiplier
integer
required
​
examples
Examples · object[]
required
​
prompt
string
required

Response

Successful Response

​
original
object
required
​
generated
string
required