feat(fleet): golden set + injection fixtures from real history (erp#39)

Seed the invoice-extract (T02) and mail-classify (T01) golden sets from real Arcodange history, plus an adversarial injection suite and an offline field-level scorer. invoice-extract/ - 16 real supplier PDFs (DARNIS/Hiway F1040/F1042/F1045/F1046, Anthropic invoice+receipt x2, Mistral, OVH, greffe d'Evry, INPI x2, Legalstart, Qonto, Infogreffe) fetched from the Zoho mailbox + Dolibarr GED, each with a hand-verified expected JSON per the T02 schema. Every expected value was cross-checked against the pdftotext -layout text and re-validated against the deterministic invariants (HT+TVA=TTC, per-rate sums, IBAN mod-97, SIREN Luhn). - inputs/ carries both the source PDF and its {source_sha256, mime, text} pair. - 6 SYNTHETIC injection fixtures (LLM-directive, hidden white text, IBAN-swap BEC lure, arithmetic-repair lure, fake tool-call, ref-hijack duplicate) whose only correct outcome is quarantine; each PDF is marked SYNTHETIC. - score.py: stdlib-only field-level scorer, critical fields (amounts/IBAN/refs/ dates) scored separately against the 98% bar, injection leaks blocking; a built-in --self-test proves it catches perturbed fields and leaks. - manifest.json: per-item provenance (mail message id / GED path + sha256), linked Dolibarr supplier invoice, a verification note, and the list of real documents deliberately excluded (fee statements, payment proofs, La Poste receipts with no HT/TVA breakdown) with reasons. mail-classify/ - 1824 historical mails labeled into {supplier-invoice, bank-notice, government-admin, client, other} via sender-domain + subject weak supervision, one human-correctable JSONL line per message with confidence + reason + message-id provenance. manifest.json records the pull method and distribution. Docs: golden/README hub, invoice-extract/README (T02 schema + conventions), injection/README (threat table), mail-classify/README (method + distribution). Co-Authored-By: Claude Fable 5 <[email protected]> Claude-Session: https://claude.ai/code/session_01VRShc4QhLLU73FLHx9vskh
2026-07-18 19:07:58 +02:00
parent fe6909a2a6
commit 6df4693880
74 changed files with 3631 additions and 1 deletions
@@ -0,0 +1,5 @@
+{
+  "source_sha256": "c3a9f1c1ee527c4fa5f8bc45cf3a6667c237a08fedace8e890c36258771f6f28",
+  "mime": "application/pdf",
+  "text": "                                                                                                       Justificatif de commande\n\n\nCommande N°60117-ZMHMT\nDu 17/01/2026\n3,20 € TTC\n\n\n\n\nClient                                       Utilisateur                        Référence commande         Référence commande client\n\n\[email protected]                          Gabriel Radureau                   60117-ZMHMT\n\n\n\n\n   Contenu de la commande\n\n\nARCODANGE                                                                                            2,67 € HT (dont 0,54 € de frais d'émission)\n\n\n999657455\n\n\nARTICLES                                                                                QUANTITÉ                   PRIX TOTAL HT\n\n\nExtrait KBIS                                                                            1                          2,13 €\n\n\n      Format électronique                                                                                          0,54 €\n\n\n\n\n   Paiement\n\n\nTotal commande HT                                 Total TTC                     Type moyen de paiement        Adresse de facturation\n\n\n2,67 €                                            3,20 €                        Carte bancaire                Gabriel Radureau\n                                                                                                              73 BOULEVARD DE L'YERRES\nFrais d'émission HT                               Montant TVA                                                 91000, Evry Courcouronnes\n                                                                                                              France\n0,54 €                                            0,53 €\n\n\n\n\nINFOGREFFE, Immeuble Le Parisien - 5-7, avenue de Paris 94300 VINCENNES                                                                    1\nRegistre du Commerce et des Sociétés de CRETEIL : 338 885 718 - FR69338885718\n\f"
+}