How sources work
What a data source is in Adduce, how the add-source wizard connects one, and how syncing, sync history and deleting a source behave.
A source is one connection between a project and a tool you already run, and every piece of evidence Adduce holds arrived through one.
A source belongs to one project
Sources are scoped to a project, never to the workspace. Two projects reading the same PostHog account each hold their own source, with their own credential and their own sync. A project may hold several sources of the same kind — two Intercom workspaces, three transcript collections — and nothing crosses between projects.
Every source has a Type from the catalog, a name you choose, and a status: Connected, Error, or Not configured. Each source shows a signal count: the pieces of evidence Adduce has taken from it.
Adding one takes four steps, or three for uploads
Add source opens a wizard. For a tool Adduce connects to over its API the steps are Choose integration, Connection details, Establishing connection and Result. For an upload source — a transcripts collection — there is nothing to connect to, so the third step is skipped and the steps are Choose integration, Name the collection and Result.
The first step lists every connector. A tile marked Beta works but has not been proven against a real customer account yet; a tile marked Coming soon cannot be chosen.
The connection test runs before anything is saved
Choosing Continue on Connection details tests what you entered before the source exists. Most connectors are checked against the tool itself: Adduce calls a cheap read endpoint with your credential, so a wrong region, a revoked token or a channel the app was never invited to is caught on the spot. A few are checked only for shape — PostHog is one — and for those the first sync is what proves access.
If the test fails, no source is created. The result names what went wrong, and the field to correct.
The first sync backfills, then Adduce syncs daily
The first sync reaches back as far as that connector's backfill window — full history for PostHog, ninety days for Intercom, thirty for Slack, a trailing window for the observability tools. Each source page names its own.
After that every source syncs once a day, at 03:00 UTC. Sync now on the source page runs one immediately. Only one sync per source runs at a time, so a long backfill is never overtaken by the nightly run.
Sync history is the record of what happened
The source page lists its recent runs, newest first, each with its outcome and how long it took. A completed run shows how much evidence it added. A failed run shows a short reason — authentication expired, rate limited by the provider, a configuration problem — and expands to the full message the tool returned.
Credentials are replaced by deleting and adding again
There is no edit form for a source's credentials. To rotate a key, delete the source and add it again with the new one; the next sync is a full backfill.
Deleting asks you to type the source's own name, and it removes the evidence that source produced along with it. Themes built from that evidence lose it too.
Where to see this in Adduce
- Sources in a project's sidebar holds the list, the Add source button, and the destinations Adduce writes to below it.
- A source's own page holds Details, Connection, Sync with its history, and the Danger zone.
- Pipeline shows how much evidence arrived, and from which source.