usage.md

Python usage

Use Subfork as a context manager to close its HTTP connection when finished. Responses are ordinary dictionaries and lists.

from subfork import Subfork

with Subfork() as client:
    graphs = client.graphs.list()
    nodes = client.nodes.list()

Create and publish

Load a graph definition from JSON, validate it, and save a new graph:

import json
from subfork import Subfork

with open("graph.json", encoding="utf-8") as stream:
    definition = json.load(stream)

with Subfork() as client:
    validation = client.graphs.validate(definition)
    if not validation.get("valid", False):
        raise ValueError(validation)
    graph = client.graphs.create(name="My graph", definition=definition)
    published = client.graphs.publish(graph["id"], version="v1")

Graphs are public. Publication creates an immutable version. To replace draft content, use client.graphs.update(graph_id, name=..., definition=..., description=...). Supply the description you want to retain; omitting it clears the description.

Execute and wait

with Subfork() as client:
    execution = client.graphs.execute(graph_id, version="v1", inputs={})
    result = client.executions.wait(execution["id"], timeout=120)
    if result["status"] == "completed":
        print(result["outputs"])
    else:
        print("Execution ended:", result["status"])

Replace graph_id with your graph ID and inputs with its named inputs. Omit version to run the draft. Runs consume your account's execution quota.

execute() submits immediately; wait() polls until a terminal state. You can also call executions.get(execution_id) or executions.cancel(execution_id). A wait timeout stops polling without canceling the remote run.

Discover reusable graphs

with Subfork() as client:
    catalog = client.graphs.published()
    block = client.graphs.published_version(graph_id, "v1")
    interface = client.graphs.interface(owned_graph_id)

Published graph versions include a composite node manifest for use in larger graphs. Pin versions to keep later publications from changing your graph. See examples for complete definitions.

Handle errors

from subfork import APIError, ExecutionTimeout, Subfork

try:
    with Subfork() as client:
        execution = client.graphs.execute(graph_id)
        result = client.executions.wait(execution["id"])
except APIError as error:
    print(error.status_code, str(error))
except ExecutionTimeout as error:
    print("Still waiting for:", error.execution_id)

API errors include AuthenticationError, PermissionDeniedError, NotFoundError, ValidationError, and RateLimitError. The client does not automatically retry requests. After a TransportError, check remote state before submitting another create, publish, or execute request.

See troubleshooting for common errors.

See the API and file workflow guide for asset uploads, artifact downloads, and using execution results in your own functions.