The monitoring engine where a new market is a config file
A watch-everything intelligence engine for one geographic market: public sources in, deduplicated and relevance-scored items out, a cited morning digest by 7 a.m. Adding a second market took a JSON file and zero code changes.
Keeping up with everything that happens in one geographic market sounds simple until you list the sources: government sites, news outlets, weather alerts, school systems, sports schedules, community calendars. No single feed covers it, and checking them all daily is a part-time job nobody actually does. So the coverage gets spotty, and the important item surfaces three days late.
What the system does
The engine is deliberately generic. A market is defined entirely by one JSON profile: its sources, its place names, its scoring hints. The pipeline reads that profile, fetches every source on a schedule, deduplicates against everything already seen, scores each item for local relevance, and files it into a full-text-searchable corpus. Every morning at 7:00 it produces a digest of what matters, with every item attributed to its source.
In its first month it ingested about 2,400 items and produced 20 daily digests, with the fetch jobs succeeding 97 percent of the time. A downstream job turns the digest into a draft morning briefing that a human reviews and publishes.
The guardrails
- Public sources only. The fetcher respects robots.txt and never goes behind a paywall or a login. If a source does not want automated readers, it is not a source.
- The AI cannot invent a story. The digest layer cites items strictly by database reference. A deterministic validator, not the model, owns every link and every claim of existence. A hallucinated item has no id and dies at validation.
- Humans publish. The engine drafts; it has no path to putting words in front of an audience by itself.
The proof that matters
The claim that something is "config-driven" is cheap until you test it. I generated a second market profile with the built-in scaffold command: it produced a working skeleton with zero code changes. Same engine, different JSON, new coverage area. That is the difference between a tool and a product: the second customer costs a config file, not a rebuild.
What this means for your business
Swap "county" for whatever you need to watch: your industry's regulators, your competitors' announcements, your supply chain's press releases, permits filed in your service area. The architecture is identical. If someone in your company spends an hour a day checking the same twenty websites, that job is a JSON file now.
The specific market and source list are withheld. The numbers are real. If there is a landscape your business needs to watch, start with a roadmap session.
