chassis-cloud 0.1.4

Official Python SDK for the Chassis GPU cloud API by OkeyMeta Ltd

chassis-cloud (Python)

Official Python client for the Chassis public API (/api/v1).

Default base URL: https://chassis.okeymeta.com.ng/api/v1

PyPI: chassis-cloud · import as chassis

Install

pip install chassis-cloud

Requires Python 3.10+ and httpx.

from chassis import Chassis

Auth

Create an API key in the Chassis console under API Keys. Keys look like chs_… and are shown once.

from chassis import Chassis

with Chassis(api_key="chs_...") as client:
    ...

Quickstart

from chassis import Chassis

with Chassis(api_key="chs_...") as client:
    gpus = client.list_gpus()
    instance = client.spin_up(
        gpuSkuId=gpus[0]["id"],
        name="gpu-host-01",
        imageName="ghcr.io/YOUR_ORG/your-gpu-app:latest",
        ports="8080/http,22/tcp",
    )

Example: host a GPU service

Any CUDA app — APIs, media tools, notebooks, batch workers — not only training.

with Chassis(api_key="chs_...") as client:
    gpus = client.list_gpus()
    instance = client.spin_up(
        gpuSkuId=gpus[0]["id"],
        name="gpu-host-01",
        imageName="ghcr.io/YOUR_ORG/your-gpu-app:latest",
        containerDiskGb=50,
        ports="8080/http,22/tcp",
        env={"MODEL_ID": "your-model"},
    )
    detail = client.get_instance(instance["id"])
    print(detail.get("publicIp"), detail.get("connection"), detail.get("status"))
    # Point your clients at publicIp / published ports
    client.stop(instance["id"])

Example: training job

Spin up a dedicated GPU, run your trainer on the instance, then stop billing.

from chassis import Chassis

with Chassis(api_key="chs_...") as client:
    gpus = client.list_gpus()
    gpu = next(
        (g for g in gpus if "A100" in str(g.get("displayName", ""))),
        gpus[0],
    )

    instance = client.spin_up(
        gpuSkuId=gpu["id"],
        name="finetune-bert",
        gpuCount=1,
        imageName="pytorch/pytorch:2.1.0-cuda11.8-cudnn8-devel",
        containerDiskGb=80,
        ports="8888/http,22/tcp",
    )
    print("running", instance["id"], instance["status"])
    # SSH / Jupyter → run train.py on the instance
    client.stop(instance["id"])

Example: multi-node cluster

with Chassis(api_key="chs_...") as client:
    gpus = client.list_gpus()
    cluster = client.create_cluster(
        name="dist-train",
        gpuSkuId=gpus[0]["id"],
        nodeCount=4,
        gpusPerNode=1,
        imageName="pytorch/pytorch:2.1.0-cuda11.8-cudnn8-devel",
    )
    print(client.get_cluster(cluster["id"]))
    # Nodes receive CHASSIS_CLUSTER_ID, CHASSIS_NODE_RANK, CHASSIS_NODE_COUNT
    client.stop_cluster(cluster["id"])
    # client.terminate_cluster(cluster["id"])

Example: serverless endpoint

with Chassis(api_key="chs_...") as client:
    gpus = client.list_gpus()
    endpoint = client.create_endpoint(
        name="text-infer",
        gpuSkuId=gpus[0]["id"],
        workersMin=0,
        workersMax=3,
    )

    result = client.run_sync(
        endpoint["id"],
        body={
            "input": {
                "prompt": "Summarize Chassis in one sentence.",
                "max_tokens": 128,
            }
        },
    )
    print(result)

    job = client.run(endpoint["id"], body={"input": {"prompt": "hello"}})
    status = client.get_job(endpoint["id"], job["id"])
    print(status)

Instance options (create_instance / spin_up)

Field	Required	Notes
gpuSkuId	yes	Chassis SKU UUID from list_gpus()
name	yes	Instance name
gpuCount	no	Default 1, max 8
imageName	no	Container image
containerDiskGb	no	Default 50
volumeGb	no	Ephemeral volume GB
networkVolumeId	no	Chassis network volume UUID
registryCredentialId	no	Chassis registry credential UUID
cloudType	no	"SECURE" or "COMMUNITY"
ports	no	e.g. "8080/http,22/tcp"
env	no	String map passed into the container
startAfterCreate	no	Default True (spin_up forces True)

Methods

GPUs & instances

Method	HTTP
list_gpus()	GET /gpus
list_instances()	GET /instances
create_instance(**kwargs)	POST /instances
spin_up(**kwargs)	POST /instances (startAfterCreate: true)
get_instance(id)	GET /instances/:id
get_instance_logs(id, tail=None)	GET /instances/:id/logs
start(id)	POST /instances/:id/start
stop(id)	POST /instances/:id/stop
restart(id)	POST /instances/:id/restart
terminate(id)	DELETE /instances/:id

Clusters (multi-node)

Method	HTTP
list_clusters()	GET /clusters
create_cluster(**kwargs)	POST /clusters
get_cluster(id)	GET /clusters/:id
start_cluster(id)	POST /clusters/:id/start
stop_cluster(id)	POST /clusters/:id/stop
terminate_cluster(id)	DELETE /clusters/:id

Create with name, gpuSkuId, nodeCount (2–8), and optional gpusPerNode, imageName, networkVolumeId, ports. Nodes get CHASSIS_CLUSTER_ID, CHASSIS_NODE_RANK, and CHASSIS_NODE_COUNT.

Templates, volumes, registries

Method	HTTP
list_templates() / create_template(**kwargs)	/templates
list_volumes() / create_volume(**kwargs)	/volumes
list_registries() / create_registry(**kwargs)	/registries

Serverless endpoints

Method	HTTP
list_endpoints()	GET /endpoints
create_endpoint(**kwargs)	POST /endpoints
delete_endpoint(id)	DELETE /endpoints/:id
run(endpoint_id, body=None)	POST /endpoints/:id/run
run_sync(endpoint_id, body=None, wait_ms=None)	POST /endpoints/:id/runsync

Create options also accept env, interruptible, and publicIp.

Responses use data / error. Failures raise ChassisError with .status, .body, and a clear message.

Docs

Human setup guide: https://chassis.okeymeta.com.ng/docs

License

MIT · OkeyMeta Ltd