Repository navigation
APUS: a no-payment HTTP research probe for optional data-processing steps #23503
daymo-sketch
started this conversation in
Show and tell
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
I’m working on APUS, a proposed agent-resource toolbox. We’ve built a small HTTP research service to find out whether existing automated workflows would actually choose bounded processing steps before we implement more of them.
For Prefect users, the relevant proposed jobs include CSV quality checks, matching rows between two tables, PDF table extraction, document version comparison, and dataset schema drift. Each offer has a bounded result, $0.01 indicative price, input contract and a free/existing alternative.
This is a research probe, not a Prefect integration or a set of working processing tools. Selections are free and nonbinding. No file, workflow data, wallet signature or money should be sent. Choosing an offer immediately returns a non-delivery explanation. We are not claiming paid demand or useful delivery from those selections.
The automation-audience entry point is:
https://apus-demand-test.smartsage32.chatgpt.site/start/automation
A minimal read-only step for inspecting the contracts:
This function just reads the index; it does not execute or charge for a step. It can be called from an existing workflow's inspection/decision stage without adding an integration package. The explicit descriptive User-Agent is needed because the hosting edge rejects Python’s default identifier.
If an agent in an existing workflow has a matching current task, the JSON
x-flowand request schemas let it consider only relevant offers, inspect the proposed output and select one, its existing/free solution, or “none”. Please do not create work just to help our test. Mark manual choiceshumanand integration checksoperator_test; these are kept separate from self-reported agent selections. Repeated submissions count once, not as recurring demand.We retain minimal pseudonymous interaction records, not task contents; app codes and agent labels are self-reported and do not verify identity. We will use the first 14-day campaign to prioritise small real deliveries, not to declare that an unexposed idea has no market.
I’m affiliated with the project and posting on its owner’s behalf. Feedback on missing workflow steps, unclear contracts, transport friction or reasons to prefer an existing library is welcome here. No purchase request, Prefect endorsement or product-quality claim is intended.
All reactions