Custom API FAQ
Does my API need anything special?
It needs a list endpoint that returns records as JSON, page by page, and a detail endpoint for one record, both answering GET over https:// on one public address. Each record needs a stable, unique ID. See When to use it.
Can I filter on the API side instead of after the fact?
Yes: add the API's own filter as a fixed query parameter on the Connection step, like status=open. Conditions on the Records step run after the API returns each page, so a filter in the request saves requests on large lists. You can use both.
Why does Preview Matches say "Unknown" for the total?
The preview reads only the first page (up to 50 records). Only a full pass through the list can count every match.
How long does the first import take?
Each step reads up to 10 pages and adds at most 10,000 documents. Steps run about every 5 minutes once the previous step's documents are processed, so a large list is imported over several steps. Processing time depends on the documents and your embedding model.
What happens when I raise the document limit?
New matching records are added as later passes reach them, up to the new limit. Raising the limit marks the source Review required first.
Do documents disappear when their records are deleted from my API?
No. Documents stay until you delete them from Documents. The same applies when a record stops matching your conditions.
How are changed records found?
From the Last updated field in the list response when you map one; otherwise from the title, text and link fields in the list response. See How changes are detected.
What does a refresh cost?
A changed record whose document turns out identical costs 1 credit and isn't re-embedded, unless the project's embedding settings or Enhanced Sync choice changed since it was last processed. A document that did change costs the same as any sync. Re-sync on a document always processes it in full.
Can members set up custom API sources?
No. Owners and Admins set them up, edit and review them. Members can see them, and browse, re-sync and tag their documents; deleting documents is for Owners and Admins.
Can I move a source to another project?
No. A source stays in the project it was connected in. Connect a new source in the other project.
Can I change the API address, the authentication type or the record ID?
No. Those fields identify the source; connect a new source instead. See What can't change later.
Where are my credentials kept?
Encrypted with the source. They're sent from your browser once, when you test the connection, and never shown again. OAuth access tokens are cached encrypted and renewed automatically.
My API is on a private network. Can I connect it?
Not on the hosted app: it won't call private or reserved addresses, or plain http://. A self-hosted deployment's operator can allow specific origins. See Allowed origins.
My API's next-page link points to a different host. Will it work?
No. Every request must stay on the API address; a link elsewhere stops the sync with a pagination error.
Can I pause a source?
Yes. Open Monitor on the source and choose Pause scanning (Owners and Admins). Scheduled checks stop where they are, and documents already queued still process. Resume scanning continues from the same place. See Monitoring and scheduled checks.