Extend trajectories with a next user turn (or estimate cost)

POST/api/v1/datasets/{dataset_id}/extend-trajectories

For each row’s conversation, generates the user’s next message: a single question or request that can only be answered correctly by remembering details from earlier in the conversation, written in the conversation’s language. No completion is generated for it. Each request creates a new dataset for the result, returned as dataset_id; the dataset in the path is left unchanged. Poll GET /datasets/{dataset_id}/status and download from GET /datasets/{dataset_id}/download with the new id. If the dataset has been adapted (POST /datasets/{dataset_id}/adapt finished), the extension builds on its adapted download: column_mapping names columns of that download, so column_mapping.completion can be enhanced_completion, and the result keeps the adapted columns next to extended_prompt. Only prompt + completion applies there, since adapt splits a chat column into original_prompt and original_completion. Otherwise the extension reads the uploaded file, with one of two mappings. column_mapping.chat: a column of {"role", "content"} turns (or a JSON string of that list), ending on the assistant’s reply. Or column_mapping.prompt with column_mapping.completion: the conversation up to the user’s last turn (a turn list, or a plain string), plus the final reply as plain text. Each row of the result carries the conversation (split as for /adapt) plus an extended_prompt column. context, universal_prompt and image mappings are not currently supported. Credits and status polling behave as for /adapt; an idempotency_key replays the extension it first created, returning its dataset_id and run_id with zero estimates, since the replay starts and charges nothing (as for /translate). Set estimate=true to validate and get a cost quote without starting a run.

Path ParametersExpand Collapse

dataset_id: string

Body ParametersJSONExpand Collapse

column_mapping: optional object { chat, prompt, completion }

Column role assignments. Required for real runs, optional for estimate-only requests.

chat: optional string

Column holding each row’s conversation as a list of {"role", "content"} turns (or a JSON string encoding that list), ending on the assistant’s reply. The full conversation is the trajectory the next user message is generated from. Send chat alone, or prompt with completion instead. Not available on an adapted dataset: adapt splits the chat column into original_prompt and original_completion, so map those with prompt and completion.

prompt: optional string

Column holding the conversation up to and including the user’s last turn: a list of {"role", "content"} turns (or a JSON string encoding that list), or a plain string read as a single user turn. Requires completion; not allowed with chat. On an adapted dataset, a column of its download: original_prompt if it was adapted from a chat column, otherwise the dataset’s own prompt column.

completion: optional string

Column holding the assistant’s final reply to prompt, as plain text. It becomes the last turn of the trajectory. Requires prompt; not allowed with chat. On an adapted dataset, a column of its download: enhanced_completion extends the adapted reply, the original completion column extends the reply as uploaded.

job_specification: optional object { max_rows, idempotency_key }

Job execution parameters

max_rows: optional number

Maximum number of rows to process in this run

minimum1

idempotency_key: optional string

Client-generated idempotency key for safe retries. If a launch with the same key already exists, the original response is returned.

estimate: optional boolean

When true, validates the request and returns the estimated credit cost without starting a run.

ReturnsExpand Collapse

run_id: optional string

Unique identifier for this pipeline run. Null for estimate-only requests.

estimatedMinutes: number

Estimated processing time in minutes

estimatedCreditsConsumed: number

Estimated number of credits that will be consumed by this run

estimate: boolean

Whether this was an estimate-only request (no run started)

multimodalPricingApplied: boolean

True when an image column is mapped and also listed in context_columns; each output row is billed at a higher rate.

creditMultiplier: optional number

10 credits per 100 output rows when multimodalPricingApplied is true. Omitted for text-only pricing.

dataset_id: string

The new dataset holding the extended trajectories. Poll GET /datasets/{dataset_id}/status and download from GET /datasets/{dataset_id}/download with this id, not the source dataset’s. Null for estimate-only requests.

formatuuid

Extend trajectories with a next user turn (or estimate cost)

curl https://api.prod.adaptionlabs.ai/api/v1/datasets/$DATASET_ID/extend-trajectories \
    -H 'Content-Type: application/json' \
    -H "Authorization: Bearer $ADAPTION_API_KEY" \
    -d '{}'

{
  "run_id": "dataset-550e8400-e29b-41d4-a716-446655440000-1712234567890",
  "estimatedMinutes": 0,
  "estimatedCreditsConsumed": 0,
  "estimate": true,
  "multimodalPricingApplied": true,
  "creditMultiplier": 0,
  "dataset_id": "7d1f3c2a-9b4e-4f60-8a1d-2e5c6b7a8f90"
}

Returns Examples

{
  "run_id": "dataset-550e8400-e29b-41d4-a716-446655440000-1712234567890",
  "estimatedMinutes": 0,
  "estimatedCreditsConsumed": 0,
  "estimate": true,
  "multimodalPricingApplied": true,
  "creditMultiplier": 0,
  "dataset_id": "7d1f3c2a-9b4e-4f60-8a1d-2e5c6b7a8f90"
}