API Reference#

Base URL#

When you run the OpenFold2 NIM locally, the API is served from the host and port you expose with docker run (the default is port 8000):

http://localhost:8000

All endpoint paths below are relative to this base URL. If you set a custom port with NIM_HTTP_API_PORT (see Configuration Alternatives at Runtime), substitute that port for 8000.

Authentication#

The self-hosted OpenFold2 NIM does not require an authentication header for inference or management requests—the container is expected to run inside your own trusted network. Requests can be sent directly, for example:

curl -X POST 'http://localhost:8000/biology/openfold/openfold2/predict-structure-from-msa-and-template' \
     -H 'Content-Type: application/json' \
     -d @request.json

An NGC_API_KEY is required only to pull the NIM container image from NGC (see Prerequisites), not to call the running API.

Endpoints#

Endpoint	Method	Description
`/biology/openfold/openfold2/predict-structure-from-msa-and-template`	POST	Predict structure from an input MSA and optional templates.
`/v1/health/live`	GET	Liveness probe. Returns 200 when the container is running.
`/v1/health/ready`	GET	Readiness probe. Returns 200 when the model is loaded and ready to serve.
`/v2/health/ready`	GET	Readiness probe (v2).
`/v1/metadata`	GET	Model and NIM metadata.
`/v1/manifest`	GET	Model manifest details.
`/v1/license`	GET	License information.
`/v1/metrics`	GET	Prometheus-format metrics.

OpenAPI Schema#

The OpenAPI specification details the full request and response schema. If the embedded reference below does not render in your environment, download the raw specification directly: openapi.json.