An ingestion pipeline independent of traffic
A persisted publication watermark and chunked D1 upserts support incremental collection. Bounded detail enrichment runs with ingestion, keeping source API work separate from visitors and crawlers.
DATA PRODUCTS / FUNDING DISCOVERY
Find the opportunity. Keep the deadline.
Public-grant discovery built around relevant opportunities: official-source ingestion, regional and sector filters, translated information and deadline reminders.
Subvenio
ReleasedDSL4-owned grant-discovery product. Official source records remain authoritative for eligibility and deadlines.
THE PRODUCT
What it solves.
Public funding information is useful only when someone can find the relevant call and act in time. Subvenio turns BDNS grant records into structured pages and filtered digests, with reminders ahead of approaching deadlines.
THE ENGINEERING
How it works.
A Cloudflare Worker ingests BDNS records on Cron Triggers, classifies them deterministically and stores them in D1. Page requests read stored data rather than triggering upstream API calls. The translation pipeline includes a Cloudflare Workers AI adapter alongside external providers, with fallback handling and usage bookkeeping for monthly-capped providers.
CLOUDFLARE & PRODUCT ARCHITECTURE
Inside the system.
Translation with explicit limits
The provider chain implements fallback between translation services, including Workers AI. D1-backed character accounting helps avoid exceeding configured budgets for monthly-capped providers; integration support does not imply every provider is used on every run.
Reminders that remember
Subscriber filters, alert cursors and recorded reminder deliveries connect collection to useful notifications. Scheduled T-14 and T-3 reminders are deduplicated so repeated processing does not intentionally resend the same deadline notice.