Run Foliqant from your application

Foliqant runs inside your Python process. Your application supplies one input, awaits a workflow, and receives one structured result. The workflow configuration chooses its steps and routes; your application decides when to call it and what to do with the result.

Run one request in Python

This example uses a configured workflow named demo in config/settings.yaml. Install and run shows how to create that configuration. Install the provider extra required by your model profile before opening the application.

import asyncio
import os
from pathlib import Path

from foliqant import Envelope, open_application, prepare_application


async def main() -> None:
    prepared = prepare_application(Path("config/settings.yaml"))
    async with open_application(prepared, environment=os.environ) as app:
        result = await app.run("demo", Envelope(payload={"message": "Please send my statement."}))
        if result.execution.status == "completed":
            print(result.payload)
        elif result.execution.status == "needs_review":
            print("A person or application policy must review this request.")
        else:
            print(result.execution.error)


asyncio.run(main())

prepare_application compiles the local configuration without opening clients. open_application resolves marked environment values and owns client startup and shutdown. Keep that context open for the lifetime of a server or worker and call app.run(...) for each request. Each call has its own execution state; the compiled plans and clients can be shared. Do not prepare and open again for every HTTP request.

The Envelope.payload is your business input. Envelope.metadata is optional context. The workflow's input schema validates the payload at admission. To decode untrusted HTTP bytes, use decode_envelope, which bounds and validates the JSON input; see Expose an HTTP endpoint.

Choose the next path

You need to…	Read
Interpret the returned business value and decision evidence	Read results
Add an HTTP route around the same application	Expose an HTTP endpoint
Distinguish review from technical failure	Handle errors
Process several independent requests from one message	Process multiple requests
Keep an application open in a service	Deploy and operate
Inspect safe runtime signals	Observe a running application

For exact Python and JSON fields, see the input and result reference.