Create training job
const url = 'https://example.com/api/v1/org/example/ws/example/training/jobs';const options = { method: 'POST', headers: {Authorization: 'Bearer <token>', 'Content-Type': 'application/json'}, body: '{"backend":"tinker","capability_ref":{"name":"example","version":"example"},"config":{"batch_size":1,"checkpoint_interval":1,"dataset_ref":{"name":"example","version":"example"},"epochs":1,"eval_dataset_ref":{"name":"example","version":"example"},"gradient_accumulation_steps":1,"learning_rate":1,"lora_alpha":1,"lora_rank":1,"max_sequence_length":1,"steps":1,"trajectory_dataset_refs":[{"name":"example","version":"example"}]},"model":"example","name":"example","project_ref":"example","run_ref":"example","tags":["example"],"trainer_type":"sft"}'};
try { const response = await fetch(url, options); const data = await response.json(); console.log(data);} catch (error) { console.error(error);}curl --request POST \ --url https://example.com/api/v1/org/example/ws/example/training/jobs \ --header 'Authorization: Bearer <token>' \ --header 'Content-Type: application/json' \ --data '{ "backend": "tinker", "capability_ref": { "name": "example", "version": "example" }, "config": { "batch_size": 1, "checkpoint_interval": 1, "dataset_ref": { "name": "example", "version": "example" }, "epochs": 1, "eval_dataset_ref": { "name": "example", "version": "example" }, "gradient_accumulation_steps": 1, "learning_rate": 1, "lora_alpha": 1, "lora_rank": 1, "max_sequence_length": 1, "steps": 1, "trajectory_dataset_refs": [ { "name": "example", "version": "example" } ] }, "model": "example", "name": "example", "project_ref": "example", "run_ref": "example", "tags": [ "example" ], "trainer_type": "sft" }'Authorizations
Section titled “Authorizations ”Parameters
Section titled “ Parameters ”Path Parameters
Section titled “Path Parameters ”Organization slug
Organization slug
Workspace slug
Workspace slug
Request Body required
Section titled “Request Body required ”Create a hosted Tinker SFT job.
object
Capability artifact used as the training policy scaffold
object
Capability artifact name
Explicit capability version
Configuration for a hosted Tinker SFT job.
object
Stable reference to a versioned dataset artifact.
object
Dataset artifact name
Explicit dataset version
Stable reference to a versioned dataset artifact.
object
Dataset artifact name
Explicit dataset version
Optional Worlds trajectory dataset references used for SFT ETL
Stable reference to a versioned dataset artifact.
object
Dataset artifact name
Explicit dataset version
Base model or adapter target
Optional job tags
Create a hosted Tinker RL job.
object
Capability artifact used as the training policy scaffold
object
Capability artifact name
Explicit capability version
Configuration for a hosted Tinker RL job.
object
Tinker RL algorithm
Stable reference to a versioned dataset artifact.
object
Dataset artifact name
Explicit dataset version
RL execution mode. one_step_off_async keeps one rollout group in flight; fully_async widens the same pipeline to bounded multi-group async training.
Stable reference to a versioned dataset artifact.
object
Dataset artifact name
Explicit dataset version
Named reward or verification recipe resolved server-side.
object
Server-side reward recipe name
Recipe parameters resolved by the training worker
object
Optional Worlds trajectory dataset references used for offline RL ETL
Stable reference to a versioned dataset artifact.
object
Dataset artifact name
Explicit dataset version
SDK-side reward policy for live Worlds RL rollouts.
object
Reward policy or preset name
Optional reward-policy parameters or component definitions
object
Base model or adapter target
Optional job tags
Create a hosted Ray GRPO job.
object
Capability artifact used as the training policy scaffold
object
Capability artifact name
Explicit capability version
Configuration for a hosted Ray GRPO job.
object
Ray RL algorithm
Ray execution mode
Prompt dataset used to seed RL episodes
object
Dataset artifact name
Explicit dataset version
Named reward or verification recipe resolved server-side.
object
Server-side reward recipe name
Recipe parameters resolved by the training worker
object
Task/environment definition reference
Base model or adapter target
Optional job tags
Responses
Section titled “ Responses ”Successful Response
Training job detail returned by hosted training endpoints.
object
object
The raw submit-time job configuration (TinkerSFTJobConfig / TinkerRLJobConfig / RayGRPOJobConfig serialized as JSON). The SDK-side fetch_and_build_payload reads fields like learning_rate, batch_size, steps from here to reconstruct the training payload in the sandbox.
object
Small denormalized user identity for cross-domain byline display.
Stable reference to a versioned dataset artifact.
object
Dataset artifact name
Explicit dataset version
object
Stable reference to a versioned dataset artifact.
object
Dataset artifact name
Explicit dataset version
Stable reference to a versioned dataset artifact.
object
Dataset artifact name
Explicit dataset version
Example
{ "algorithm": "grpo", "backend": "tinker", "status": "queued", "trainer_type": "sft"}Invalid request
Canonical API error envelope returned by the API.
object
HTTP status code
Human-readable error message
object
Where the validation error occurred
Human-readable validation message
Machine-readable validation error type
Stable machine-readable error type
Example
{ "code": 401, "detail": "Authentication failed", "type": "authentication_error"}Authentication failed
Canonical API error envelope returned by the API.
object
HTTP status code
Human-readable error message
object
Where the validation error occurred
Human-readable validation message
Machine-readable validation error type
Stable machine-readable error type
Example
{ "code": 401, "detail": "Authentication failed", "type": "authentication_error"}Access forbidden
Canonical API error envelope returned by the API.
object
HTTP status code
Human-readable error message
object
Where the validation error occurred
Human-readable validation message
Machine-readable validation error type
Stable machine-readable error type
Example
{ "code": 401, "detail": "Authentication failed", "type": "authentication_error"}Not found
Canonical API error envelope returned by the API.
object
HTTP status code
Human-readable error message
object
Where the validation error occurred
Human-readable validation message
Machine-readable validation error type
Stable machine-readable error type
Example
{ "code": 401, "detail": "Authentication failed", "type": "authentication_error"}Already exists
Canonical API error envelope returned by the API.
object
HTTP status code
Human-readable error message
object
Where the validation error occurred
Human-readable validation message
Machine-readable validation error type
Stable machine-readable error type
Example
{ "code": 401, "detail": "Authentication failed", "type": "authentication_error"}Validation error
Canonical API error envelope returned by the API.
object
HTTP status code
Human-readable error message
object
Where the validation error occurred
Human-readable validation message
Machine-readable validation error type
Stable machine-readable error type
Example
{ "code": 401, "detail": "Authentication failed", "type": "authentication_error"}Upgrade required
Canonical API error envelope returned by the API.
object
HTTP status code
Human-readable error message
object
Where the validation error occurred
Human-readable validation message
Machine-readable validation error type
Stable machine-readable error type
Example
{ "code": 401, "detail": "Authentication failed", "type": "authentication_error"}Rate limited
Canonical API error envelope returned by the API.
object
HTTP status code
Human-readable error message
object
Where the validation error occurred
Human-readable validation message
Machine-readable validation error type
Stable machine-readable error type
Example
{ "code": 401, "detail": "Authentication failed", "type": "authentication_error"}Internal server error
Canonical API error envelope returned by the API.
object
HTTP status code
Human-readable error message
object
Where the validation error occurred
Human-readable validation message
Machine-readable validation error type
Stable machine-readable error type
Example
{ "code": 401, "detail": "Authentication failed", "type": "authentication_error"}Bad gateway
Canonical API error envelope returned by the API.
object
HTTP status code
Human-readable error message
object
Where the validation error occurred
Human-readable validation message
Machine-readable validation error type
Stable machine-readable error type
Example
{ "code": 401, "detail": "Authentication failed", "type": "authentication_error"}Gateway timeout
Canonical API error envelope returned by the API.
object
HTTP status code
Human-readable error message
object
Where the validation error occurred
Human-readable validation message
Machine-readable validation error type
Stable machine-readable error type
Example
{ "code": 401, "detail": "Authentication failed", "type": "authentication_error"}