All twelve streams in one record shape, so you write one parser instead of twelve. Run it with no input to see it work. The single-source Actors in this account each return their own source's shape; this one returns every stream in the same envelope, with the score inputs, the formula version and a proof reference on every record.
What it does
- One row per event, identical across all twelve streams.
- Filter by stream, by ticker, or take everything in one call.
- Cursor-based: pass back the cursor from your last run and you get only what is new, with no duplicates and no gaps.
- Pay per event delivered, so a run that returns nothing costs nothing.
The twelve streams
Insider buying clusters, planned insider sales (Form 144), activist stakes (13D/G), institutional holdings (13F), material corporate events (8-K) and private capital raises (Form D) from SEC EDGAR. Drug approvals and regulatory actions from openFDA. Trial readouts from ClinicalTrials.gov. Research grants from NIH RePORTER. Federal contract awards from USAspending.gov. Congressional trades from the US House Clerk. Hiring signals read from companies' own career sites.
What a record carries
event_type, company, occurred_at, filed_at, source, data with the fields specific to that stream, score, score_label, score_inputs, scored_on, proof and _cursor.
If an input is missing from the published source, the record says so rather than guessing. A score you cannot recompute is a number you cannot defend.
How it differs from the single-source Actors
The individual Actors read one source each and are the right choice if you only want that one. This Actor is the join: same envelope for every stream, one cursor across all of them, and the proof reference looked up in the same published hash chain.
Honest limits
Not real time. Sources refresh once a night, congressional trades weekly, institutional holdings quarterly. The run log prints the delay on the feed it used.
The upstream service sleeps when nobody is using it and answers starting_up once; this Actor retries that by itself, so you only notice it as a slower first run.