Is it possible to start a run just using curl?

I believe it is, but I don’t know the exact format. This is the Python request version which should get you started:

query = """query {flow{ name, id}}"""


url = '<https://api.prefect.io>'
r = <http://requests.post|requests.post>(url, json={'query': query}, headers={"authorization": "Bearer MY_TOKEN_HERE"})
print(r.status_code)
print(r.text)

@Alvaro Durán Tovar given that you asked about curl, it seems like you want to trigger a flow run using the CLI. What would be even easier than curl would be to leverage the Prefect CLI. If your flow is already registered, you could use:

prefect run --name your-unique-flow-name --watch

The extra --watch flag allows you to see the flow run logs directly in the terminal. If your flow is not registered, you can accomplish the same providing a path:

prefect run -p /path/to/flow.py

I know, I'm just trying to find a way to setup a script with minimal requirements to trigger runs

Can you share more about your use case? How would you like this script to be triggered - via CLI command, API call? Where would it run? Is it something you would trigger based on some event, or should it run on schedule?

Sure So I want to improve our UX when using metaflow, with kubernetes not with batch. Metaflow doesn't have a detached mode which means you must keep your computer open otherwise the master process dies. I want to have a script to run metaflow flows from git repos, given the correct config trigger the runs on kubernetes. This is already running nicely, giving a repo URL, commit and params starts a pod on k8s running metaflow on the cloud instead of from the developer computer.

@Alvaro Durán Tovar this should work with curl - provide your flow ID and API key:

curl '<https://api.prefect.io>' \
  -X POST \
  -H 'content-type: application/json' \
	-H 'authorization: Bearer XXX' \
  --data '{
    "query":"mutation {create_flow_run(input: { flow_id: \"05f1872f-44b3-443a-bbd4-0a1198351cad\" }) {id}}"
  }'

Nice thanks

Yeah you need it because that mutation specifically takes a flow_id. Seems like you’d need to query using the

flow

route

In that case, you’ll likely won’t be using curl for this right? Probably a Python script with requests would be easier:

import requests

query = """
 mutation {
  create_flow_run(input: { flow_id: "e9827ce5-f686-4183-b3e9-c629d63d7510" }) {
    id
  }
}
"""

url = "<https://api.prefect.io>"
response = <http://requests.post|requests.post>(
    url, json={"query": query}, headers={"authorization": "Bearer XXX"}
)
print(response.status_code)
print(response.text)

You can use the same requests syntax to query for flow ID.

👌

I think you may want to use the version_group_id instead? if a new flow version is updated, the flow_id seems to change. and I tried with flow_group_id and that didn't seem to work, but it seems to run fine ( pick up the most recent version ) using the version_group_id. Ideally create_flow_run could take project and flow name....

The

create_flow_run

task takes in project and flow name if you can install Prefect to trigger the new Flow run. This setup above is if you can’t install Prefect on the machine

right, but not by using the https graphql api. 🙁