[FEATURE REQUEST] Provide a webhook endpoint to trigger a scan

[s1rd4v3]

With paperless-ngx 2.14 it is now possible to trigger a webhook url with params incl. the document data from within a workflow docs.

I think it would be great if paperless-ai is providing a webhook route which

• accepts the document data to process
  or
• the document id which can then be fetched by paperless-ai's paperlessService.js

On the setup page, the user then can choose between a cronjob OR event based processing (via webhook) of his documents.

Optionally (since I suppose most of the paperless-ai instances are running in local network only), the webhook route should be protected with a random uid which needs to be provided when calling the endpoint to prevent unauthorized triggering. The predefined random uid should be visible from the setup page.

[clusterzx]

That's a really good idea! I like that.
Will be implemented ASAP.

Regarding the UID there is already the API Key for Paperless-AI in the Settings overview.

[MephistoJB]

I have an Addition to that. In Workflows you are able to add text when calling the endpoint. How about adding a prompt there? Then you had the absolute flexibility and it should be not much more work when adding this enhancement.

[MephistoJB]

Just a Short question. Is this Part of your latest Release?

[clusterzx]

But basically the webhook has to look like this:

[MephistoJB]

Thanks for this Feature and for the explanation. I would Like to See the Addition I mentioned above as well and I would be willing to support here by adding this feature by myself. As this repo is in hardcore and superfast changing mode I would like to avoid working on something you already implemented or planning to implement soon. Whats your idea on this?

[clusterzx]

I still struggle with the benefit of that. Can you give me some examples what will be improve by that?

Sorry for asking maybe my horizon is to small to see that 😅😅

[MephistoJB]

Though it is a cool feature to let an AI run on all the documents there are reoccuring documents like reoccuring invoices, bank account statements etc, I wish to be named and classified all the same. Like all my bankaccount statements are named like: year-mm-mm-yyyy. In Year from month to month. Than Paperlessngx put it on my fileservice in a certain format and enriched with correspondent etc. This fileservice is what I backup and share. This is just an example but I have several cases like this. Maybe you remember but a while a go I had the idea to implement a rulebased approach like in a postscript. With all the features added to paperless and to this project, this would be the last tiny part to reach that goal with this project which is so much better and more reliable than a postscript triggered by paperless.

[clusterzx]

Ok let me get my head arround that, so you would setup multiple workflows in paperless-ngx that will trigger the API Endpoint of paperless-ai with different prompts?

For example:

1. For invoices where the prompt differs
2. For contracts where you have another prompt
3. ... and so on ?

Is that what you mean?

[MephistoJB]

Exactly. And even more differentiated. For invoices coming from correspondent x another one than from y. But that's logic on paperless side.

[clusterzx]

Ok, I will implement that. :)

[MephistoJB]

Thanks a lot. That's great. You could also hand over the field you want to change. I experimented a bit with it.