Sign in

v1

latestSwagger 2.02026-08-04232286653.5 KB
Foundry

Deploys a LoRA adapter (a fine-tuned model) onto its parent base model in the project.

The deployment is asynchronous: the response returns the deployment record in a pending state, which is reconciled to ready once the adapter is loaded.

post/projects/{project_id}/foundry/loras

Path parameters

project_idstring required

ID of the project to deploy the adapter in.

Request body

model_idstring required

ID of the fine-tuned model (LoRA adapter) to deploy, as registered in the model registry.

Response

cluster_namestring

Name of the cluster the adapter is deployed to.

created_atstring

RFC 3339 timestamp when the deployment was created.

endpoint_aliasstring

Serving alias for the adapter endpoint.

evolution_uristring

Evolution artifact URI the adapter weights were fetched from.

finetuning_job_idstring

ID of the fine-tuning job that produced the adapter, if any.

idstring

Unique identifier of the LoRA adapter deployment.

last_errorstring

Most recent error message, if the deployment failed.

loader_servicestring

The loader service handling the adapter fetch/load.

model_idstring

The registry model ID of the adapter.

parent_model_namestring

Name of the parent base model the adapter is loaded onto.

project_idstring

ID of the project that owns the deployment.

rankinteger

LoRA rank of the adapter.

serving_endpoint_namestring

Internal serving endpoint name.

statusstring

Current lifecycle status of the adapter deployment. One of: pending, downloading, ready, failed, deleting, expired.

updated_atstring

RFC 3339 timestamp when the deployment was last updated.