Automate post-processing

What is this about?

Post-processing routines run automatically after the AI has created a post’s text. There are two types:

• Instructions (AI prompt): The AI revises the text according to your instruction, for example shortening it, adding SEO terms or adjusting the tone.
• Web service URL: ALLMEDIA.AI sends the text to an external web service and takes over the result.

Routines do not apply globally. They only take effect in channels to which you attach them in the channel canvas.

Prerequisites

• At least the Editor role.
• For the web service variant, your own publicly reachable service.

Step by step

1. Open post-processing. Go to Settings → Content & channels → Post-processing. The page is called “Post-Processing Routines” and has the CUSTOM ROUTINES and TEMPLATES tabs.
2. Create a routine. Click New post-processing („Neue Nachverarbeitung“) (with an empty list, Create first post-processing („Erste Nachverarbeitung erstellen“)). In the full-screen editor, enter name and description.
3. Choose the execution type. Under EXECUTION TYPE, choose either Instructions (AI Prompt) and describe under INSTRUCTIONS how the text should be edited, or Webservice URL and enter the address under WEBSERVICE.
4. Save. The routine appears as a card labeled AI (AI instruction) or API (web service).
5. Attach the routine to a channel. Open the channel’s channel canvas and select the routines you want in the Post-processing („Nachverarbeitung“) building block. They run automatically from the next post creation.

Options at a glance

• Templates: In the TEMPLATES tab you find ready-made routines. If you take one over, you get your own copy that you can edit freely.
• Card menu: Via a card’s three-dot menu you choose Duplicate or Delete. Clicking the card opens it for editing.
• Several routines: If several routines are attached to a channel, they run one after the other. Each routine works with the result of the previous one.

How the web service variant works

• ALLMEDIA.AI calls the URL via POST. The body is JSON with the field content, which contains the current text.
• Your service responds with JSON. The first available field from content, processedText, text or result is taken over.
• The request times out after 30 seconds. On a server error (5xx), it is retried once.
• If the call fails, the text remains unchanged and the generation continues.
• Addresses in private networks are blocked for security reasons. The service must be publicly reachable.

Tips

• Phrase AI instructions briefly and clearly, for example “Shorten the text to at most 120 words. Don’t change any facts.”
• Test new routines on a test channel first, before attaching them to productive channels.
• If a web service doesn’t respond, first check the URL and whether the service is publicly reachable and returns JSON.

Frequently asked questions

Can I store credentials for the web service?

No. In the routine you only enter the web service URL. If your service needs to be protected, secure it via the URL itself, for example with a secret path or token in the address.

Do routines also work on images?

No. Post-processing routines edit texts. For images, use media pipelines.

Related articles

• Channel canvas and approval workflow
• Defining content structures
• Media pipelines for automatic images