HTTP API overview | Dagnam.AI Docs

HTTP API overview The dagnam SDK and CLI are built entirely on this HTTP API, so most tasks are easier through them. This section documents the contract directly, for building your own client or calling the API from another language. Base URL and versions The API lives at https://api.dagnam.ai . Every route documented here is under /api/v1/... , except /health at the root and the OpenAI-compatible routes, which are under /v1/... see OpenAI-compatible endpoint /docs/deployments/openai-compatible . Only v1 exists today; there is no header-based versioning. Use the SDK when you can The dagnam Python package see SDK and CLI /docs/dag-lib/install is the supported client: it handles authentication, retries, idempotency, pagination and streaming for you. Reach for the raw HTTP contract only when you are integrating from another language or a tool that cannot depend on the SDK. Pagination Most list routes accept ?page= starting at 1 and ?limit= default 20, maximum 100 , and respond with: A few differ. Dataset listing and browsing page with ?skip= and ?limit= and return a plain array. Training logs and metrics and deployment logs allow a larger ?limit= . A project's version list omits pages , and deployment revisions and your model registry entries come back as a plain array. Request IDs Every response carries an X-Request-ID header. Include it when you contact Support /support about a specific call. Idempotency POST /api/v1/training/jobs , POST /api/v1/deployments and POST /api/v1/deployments/from-model-version honor an Idempotency-Key header. A repeated request with the same key returns the original response with Idempotency-Replayed: true ; the same key while the first request is still in flight answers 409; the same key with a different body answers 422. Keys are not stored across a 429 or a 5xx response, and stored results do not last forever, so use a key only to retry the same request. A replay never repeats a deployment's one-time key: api key comes back null with Idempotency-Secret-Redacted: true , so regenerate the key if you lost the first response. Size limits A request body is capped at 550 MB overall, and each uploaded dataset file at 500 MB. Above either cap the API answers 413, or 402 when your plan's per-file limit is reached first. Your plan may set a lower limit for a specific action, and a hub model file over your plan's upload limit answers 402; see Billing & usage /pay for your account's live limits.