REST API¶

The Prefect REST API is used for communicating data from clients to the Prefect server so that orchestration can be performed. This API is consumed by clients such as the Prefect Python SDK or the server dashboard.

Prefect Cloud and a locally hosted Prefect server each provide a REST API.

Prefect Cloud:
- Interactive Prefect Cloud REST API documentation
- Finding your Prefect Cloud details
Locally hosted open-source Prefect server:
- Interactive REST API documentation for a locally hosted open-source Prefect server is available at http://localhost:4200/docs or the /docs endpoint of the PREFECT_API_URL you have configured to access the server. You must have the server running with prefect server start to access the interactive documentation.
- Prefect REST API documentation

Interacting with the REST API¶

You have many options to interact with the Prefect REST API:

Create an instance of PrefectClient
Use your favorite Python HTTP library such as Requests or HTTPX
Use an HTTP library in your language of choice
Use curl from the command line

PrefectClient with Prefect server¶

This example uses PrefectClient with a locally hosted Prefect server:

import asyncio
from prefect.client import get_client

async def get_flows():
    client = get_client()
    r = await client.read_flows(limit=5)
    return r

r = asyncio.run(get_flows())

for flow in r:
    print(flow.name, flow.id)

if __name__ == "__main__":
    asyncio.run(get_flows())

Output:

cat-facts 58ed68b1-0201-4f37-adef-0ea24bd2a022
dog-facts e7c0403d-44e7-45cf-a6c8-79117b7f3766
sloth-facts 771c0574-f5bf-4f59-a69d-3be3e061a62d
capybara-facts fbadaf8b-584f-48b9-b092-07d351edd424
lemur-facts 53f710e7-3b0f-4b2f-ab6b-44934111818c

Requests with Prefect¶

This example uses the Requests library with Prefect Cloud to return the five newest artifacts.

import requests

PREFECT_API_URL="https://api.prefect.cloud/api/accounts/abc-my-cloud-account-id-is-here/workspaces/123-my-workspace-id-is-here"
PREFECT_API_KEY="123abc_my_api_key_goes_here"
data = {
    "sort": "CREATED_DESC",
    "limit": 5,
    "artifacts": {
        "key": {
            "exists_": True
        }
    }
}

headers = {"Authorization": f"Bearer {PREFECT_API_KEY}"}
endpoint = f"{PREFECT_API_URL}/artifacts/filter"

response = requests.post(endpoint, headers=headers, json=data)
assert response.status_code == 200
for artifact in response.json():
    print(artifact)

curl with Prefect Cloud¶

This example uses curl with Prefect Cloud to create a flow run:

ACCOUNT_ID="abc-my-cloud-account-id-goes-here"
WORKSPACE_ID="123-my-workspace-id-goes-here"
PREFECT_API_URL="https://api.prefect.cloud/api/accounts/$ACCOUNT_ID/workspaces/$WORKSPACE_ID"
PREFECT_API_KEY="123abc_my_api_key_goes_here"
DEPLOYMENT_ID="my_deployment_id"

curl --location --request POST "$PREFECT_API_URL/deployments/$DEPLOYMENT_ID/create_flow_run" \
  --header "Content-Type: application/json" \
  --header "Authorization: Bearer $PREFECT_API_KEY" \
  --header "X-PREFECT-API-VERSION: 0.8.4" \
  --data-raw "{}"

Note that in this example --data-raw "{}" is required and is where you can specify other aspects of the flow run such as the state. Windows users substitute ^ for \ for line multi-line commands.

Finding your Prefect Cloud details¶

When working with the Prefect Cloud REST API you will need your Account ID and often the Workspace ID for the workspace you want to interact with. You can find both IDs for a Prefect profile in the CLI with prefect profile inspect my_profile. This command will also display your Prefect API key, as shown below:

PREFECT_API_URL='https://api.prefect.cloud/api/accounts/abc-my-account-id-is-here/workspaces/123-my-workspace-id-is-here'
PREFECT_API_KEY='123abc_my_api_key_is_here'

Alternatively, view your Account ID and Workspace ID in your browser URL. For example: https://app.prefect.cloud/account/abc-my-account-id-is-here/workspaces/123-my-workspace-id-is-here.

REST Guidelines¶

The REST APIs adhere to the following guidelines:

Collection names are pluralized (for example, /flows or /runs).
We indicate variable placeholders with colons: GET /flows/:id.
We use snake case for route names: GET /task_runs.
We avoid nested resources unless there is no possibility of accessing the child resource outside the parent context. For example, we query /task_runs with a flow run filter instead of accessing /flow_runs/:id/task_runs.
The API is hosted with an /api/:version prefix that (optionally) allows versioning in the future. By convention, we treat that as part of the base URL and do not include that in API examples.
Filtering, sorting, and pagination parameters are provided in the request body of POST requests where applicable.
- Pagination parameters are limit and offset.
- Sorting is specified with a single sort parameter.
- See more information on filtering below.

HTTP verbs¶

GET, PUT and DELETE requests are always idempotent. POST and PATCH are not guaranteed to be idempotent.
GET requests cannot receive information from the request body.
POST requests can receive information from the request body.
POST /collection creates a new member of the collection.
GET /collection lists all members of the collection.
GET /collection/:id gets a specific member of the collection by ID.
DELETE /collection/:id deletes a specific member of the collection.
PUT /collection/:id creates or replaces a specific member of the collection.
PATCH /collection/:id partially updates a specific member of the collection.
POST /collection/action is how we implement non-CRUD actions. For example, to set a flow run's state, we use POST /flow_runs/:id/set_state.
POST /collection/action may also be used for read-only queries. This is to allow us to send complex arguments as body arguments (which often cannot be done via GET). Examples include POST /flow_runs/filter, POST /flow_runs/count, and POST /flow_runs/history.

Filtering¶

Objects can be filtered by providing filter criteria in the body of a POST request. When multiple criteria are specified, logical AND will be applied to the criteria.

Filter criteria are structured as follows:

{
    "objects": {
        "object_field": {
            "field_operator_": <field_value>
        }
    }
}

In this example, objects is the name of the collection to filter over (for example, flows). The collection can be either the object being queried for (flows for POST /flows/filter) or a related object (flow_runs for POST /flows/filter).

object_field is the name of the field over which to filter (name for flows). Note that some objects may have nested object fields, such as {flow_run: {state: {type: {any_: []}}}}.

field_operator_ is the operator to apply to a field when filtering. Common examples include:

any_: return objects where this field matches any of the following values.
is_null_: return objects where this field is or is not null.
eq_: return objects where this field is equal to the following value.
all_: return objects where this field matches all of the following values.
before_: return objects where this datetime field is less than or equal to the following value.
after_: return objects where this datetime field is greater than or equal to the following value.

For example, to query for flows with the tag "database" and failed flow runs, POST /flows/filter with the following request body:

{
    "flows": {
        "tags": {
            "all_": ["database"]
        }
    },
    "flow_runs": {
        "state": {
            "type": {
              "any_": ["FAILED"]
            }
        }
    }
}

OpenAPI¶

The Prefect REST API can be fully described with an OpenAPI 3.0 compliant document. OpenAPI is a standard specification for describing REST APIs.

To generate Prefect server's complete OpenAPI document, run the following commands in an interactive Python session:

from prefect.server.api.server import create_app

app = create_app()
openapi_doc = app.openapi()

This document allows you to generate your own API client, explore the API using an API inspection tool, or write tests to ensure API compliance.