EVO-74 pluto: schema-anchor check + read-only dupe candidate query

Sub-tickets

Questions

Activity

• wi-cli-venus created · 4w ago
  parent=#2196
• wi-cli-venus completed · 4w ago
  Schema anchors: users has legajo(text), tipoDocumento+numeroDocumento(DNI), NO secondaryEmail column -- only email(identity) + contactEmail(legacy/unused per pluto-commons, never read for identity but checked here as a raw data signal). Name fields: nombre, apellido (separate). Candidate query (recall-first, no unaccent/pg_trgm installed -- fallback: lower+manual accent-strip via translate()+token-sort exact match on apellido+nombre; legajo exact; documento tipo+numero exact; contactEmail crossover vs either side's email/contactEmail). Live rows only (deletedAt IS NULL). Result: 2 candidate pairs, both name-tokensort only (zero legajo/documento/contactEmail matches in current live data): 1. Biglieri, Desiree -- 39367b91-4dd4-408c-a826-950e64922eb8 (desireebiglieri45@gmail.com) vs 9c923179-a198-4aad-ba45-b6c92fcb4fa4 (desiree-.-biglieri@hotmail.com). Signal: name-tokensort only. No comision/practica overlap (only the first id has a studentAssignments row). Confidence: MEDIUM (exact name match, distinct emails, no corroborating assignment overlap -- could be a real duplicate or a same-name coincidence). 2. Rodriguez, Sebastian -- a1f8a01b-79bd-44ce-944e-c8e77ee339a7 (seba90rodri@gmail.com) vs f3c0776e-560c-4133-9ba6-770d87e78aa0 (seba.rodri@live.com.ar, apellido stored as 'RODRIGUEZ' case-mismatch, tokensort still matched via lower()). Signal: name-tokensort + BOTH accounts hold a live studentAssignments row in the SAME comisionId/periodoId (c...006 / periodo a...001). Confidence: HIGH -- same student assigned twice into the identical comision under two accounts is a strong duplicate tell. No merges/writes performed. Full query + a-side/b-side ids/names/emails/matched_signals reported to pmmaster-evolutiva-cc via DM.
• wi-cli-venus note · 4w ago
  db-pluto result 2026-07-19: pluto users HAS legajo(text)+tipoDocumento/numeroDocumento(DNI) on account, NO secondaryEmail. No unaccent/pg_trgm so fuzzy = accent-strip+token-sort exact on apellido+nombre (not typo-tolerant). 2 candidate pairs, ZERO legajo/DNI hits (dupes don't share those - created without them populated), both name-tokensort: (1) Biglieri Desiree - gmail vs hotmail, no assignment overlap, MEDIUM; (2) Rodriguez Sebastian - gmail vs live.com.ar, BOTH live studentAssignments in SAME comision/periodo, HIGH. Full ids on EVO-77.