---
title: "Start"
method: POST
path: "/api/v1/deploy"
tags: ["deployments"]
---

# Start

`POST /api/v1/deploy`

Kick off the deploy model pipeline.

## Request body

- StartDeployRequest
  - `checkpoint_id` string, required
  - `model_slug` string, required
  - `role` 'production' | 'test'
  - `output_speed` 'normal' | 'turbo' — Which GPU tier a deployment runs on: H100 (normal) or B200 (turbo).
  - `expected_concurrency` 'low' | 'medium' | 'high' — Traffic level a deployment is sized for; picks its replica count.

## Response `202`

Successful Response

- StartDeployResponse
  - `deployment_id` string, required
  - `role` 'production' | 'test', required
  - `reused_merge` boolean — True when an existing merged model was reused and the merge stage was skipped.

---

[API](https://skmtc.dev/trajectory/apis/trajectory-api.md) · [All operations](https://skmtc.dev/trajectory/apis/trajectory-api/llms.txt) · [OpenAPI document](https://skmtc-service-production.skmtc.workers.dev/v1/apis/trajectory/trajectory-api/revisions/d930b23b9602/schema)