Wealth Management 1
Pricing
Pay per usage
Pricing
Pay per usage
Rating
0.0
(0)
Developer
Tjeerd Noorland
Actor stats
0
Bookmarked
2
Total users
1
Monthly active users
7 days ago
Last modified
Categories
Share
Ingest the AFM MAR 19 (Melding artikel 19 MAR) CSV export. This is the primary source. If disabled and substantial holdings is also disabled, the run fails. If MAR 19 fails to fetch, the run fails regardless of other sources.
Ingest the AFM substantial holdings (Meldingen substantiële deelnemingen) CSV export. This is a secondary confirmation source. If this source fails after retries, the run enters degraded mode and continues with MAR 19 only. Does not cause a fatal failure on its own.
Enable the optional Exa neural search confirmation pass. When enabled, the top review records (bucket A + top N bucket B) are enriched with Exa web search results to confirm person context, role, and disposal language. Requires exaApiKey or EXA_API_KEY environment variable. If Exa is unavailable, the run continues without confirmation and records are marked accordingly.
Optional override for the AFM MAR 19 CSV download URL. Leave blank to use the default AFM endpoint. Use this to point at a fixture file or a cached copy for testing.
Optional override for the AFM substantial holdings CSV download URL. Leave blank to use the default AFM endpoint. The file is approximately 95 MB and 250,000+ rows — early lookback filtering keeps memory bounded.
Signals older than this many days are dropped during the early lookback filter step. Applies to both sources. Lower values produce fewer but more recent signals. Default 45 days covers the most recent AFM MAR 19 and substantial holdings filings.
Records with signal_confidence below this threshold are blocked by the 'below_min_signal_confidence' gate and do not reach match-ready or shortlist. Default 0.6. Lowering this value increases recall but reduces precision.
Records with natural_person_confidence below this threshold are blocked by the 'low_natural_person_confidence' gate. Default 0.6. Institutional-sounding names score near 0.1; clear person names score 0.76+.
When true, records flagged as high institutional risk (institutional_risk=high) are removed from all exports after scoring. They are still counted in the raw archive. Disable this only for diagnostic runs.
Maximum number of records written to the 'review' named dataset. Records are ranked by bucket (A → B → C), issuer desirability, NL relevance, and natural person confidence before slicing.
Maximum number of records written to the 'match-ready' named dataset. Match-ready records must pass all strict signal gates. Without Exa confirmation, match-ready output is typically 0 because records require enrichment_context or verified role data.
Maximum number of records written to the 'shortlist' named dataset. The shortlist uses a softer gate than match-ready: requires review bucket A or B, natural_person_confidence ≥ 0.45, and signal_confidence ≥ 0.40. Does not require Exa confirmation.
How many bucket B review records to include in the Exa confirmation pass (in addition to all bucket A records). Only relevant when runExaConfirmation is true. Set to 0 to confirm bucket A only.
Maximum age in hours for Exa content fetches during confirmation. Content older than this will not be fetched. Default 72 hours covers a 3-day window around the signal date. Only relevant when runExaConfirmation is true.
Optional. Secret Exa API key. If left empty, the actor will fall back to EXA_API_KEY from the environment.
