Diff
Since-timestamp polling
Pass the cursor from your last sync; get back created / updated / removed entries with the changed fields. Idempotent and resumable.
Drug Database
Capability · Change feed
Every change, since a timestamp.
Poll GET /changes?since=… for a diff of everything that moved, or register a webhook and get pushed — the watchtower turns 30+ registers into one revalidatable stream.
GET /api/v1/changes?since={ts}
What you get
Built for production drug data
The ingestion watchtower records every field-level change as it stages, validates and promotes data from 30+ public registers. You consume that as a diff: GET /api/v1/changes?since={timestamp} returns everything created, updated or removed since you last synced.
Prefer push? Register a webhook and the same events arrive at your endpoint. Either way you keep a local mirror current without re-crawling 610k rows — and you only pay for what moved.
Push
Webhooks
POST /api/v1/webhooks to subscribe, DELETE to unsubscribe. Events are signed so you can verify they came from us.
Watchtower
Schema-drift alerts
When an upstream register changes shape, the watchtower flags it within 24h and freezes the source rather than ingesting malformed data — schema_state rides every response.
Freshness
Stale-data freeze
A source that stops updating is frozen with its last_synced_at, so the feed never presents stale data as fresh.
Try it
A real call, end to end
Pass the cursor from your last sync and get back only what changed — created, updated and removed.
GET /api/v1/changes?since=…
curl 'https://drug-database.com/api/v1/changes\
?since=2026-06-26T00:00:00Z&limit=2' \
-H 'Authorization: Bearer dd_live_…'response
{
"changes": [
{
"type": "drug.updated",
"id": "3f8a1c2d-1b4e-4c7a-9d2f-0a1b2c3d4e5f",
"fields": ["is_reimbursed", "public_price"],
"source": "bag_sl",
"changed_at": "2026-06-27T02:14:00Z"
},
{
"type": "interaction.created",
"atc_a": "B01AA03",
"atc_b": "N02BA01",
"source": "epha",
"changed_at": "2026-06-27T02:15:10Z"
}
],
"next_cursor": "2026-06-27T02:15:10Z",
"_meta": { "has_more": false }
}The watchtower
- 554,523 tracked changes and counting, across drugs and interactions.
- Schema-validated staging: data is checked against an expected schema before it is ever promoted.
- Schema-drift detection within 24h; a drifting source is frozen, not silently ingested.
- Every API response carries last_synced_at and schema_state, so freshness is a field, not a guess.
Good to know
- · The changes feed needs the changes:read scope; webhooks need webhooks:write.
- · Cursors are timestamps — store next_cursor and pass it as the next since.
FAQ
Common questions
Do an initial import, then poll GET /api/v1/changes?since={cursor} on your schedule. Apply the created/updated/removed entries and store next_cursor as the since for your next call. You never re-crawl the full catalog.
Yes. Register an endpoint with POST /api/v1/webhooks and the same change events are delivered to you, signed so you can verify origin. DELETE /api/v1/webhooks removes the subscription.
It is the ingestion pipeline that stages, schema-validates and diffs every public source before promoting data. It detects upstream schema drift within 24 hours and freezes a source rather than ingest malformed data.
Every response carries last_synced_at and schema_state for the backing sources. A frozen or stale source is reported as such, so you can decide whether to trust or defer.
554,523 field-level changes so far, across drugs and interactions, and the count grows with every ingestion run.
Ready when you are
Put change feed in production this afternoon.
Spin up a key in minutes. The free tier covers prototyping and low-volume production.
Free tier · 100 requests/day · No card required