**Motivations:**
- Make certificates reproducible when CSV columns do not encode the palier
- Avoid FileNotFoundError when writing certificates into new folders
- Reuse scission in the local H6 generator to avoid duplicated certificate logic
**Root causes:**
- palier inference relied on max residue value when the class column was generic
- scission assumed output directories already exist
- empty CSV fields were coerced to 0
**Correctifs:**
- Infer palier from explicit columns (palier/m) or filename, keep heuristic fallback
- Create parent directory for output JSON
- Skip empty class/sister values instead of adding residue 0
**Evolutions:**
- Use collatz_scission for certificate generation in local H6 artefacts generator
**Pages affectées:**
- applications/collatz/collatz_k_scripts/collatz_scission.py
- applications/collatz/collatz_k_scripts/collatz_generate_local_h6_artefacts.py
- docs/fixKnowledge/collatz_scission_palier_inference_and_output_dirs.md
**Motivations:**
- Make H6(E) checks reproducible and citeable for all 60 B12 states
- Provide an aggregated view of Δm(E) and minimal observed horizons
**Root causes:**
- applications/collatz/out/ is Git-ignored, so local H6 artefacts must live in versioned docs paths
- Per-state H6 evidence was previously limited to E1
**Correctifs:**
- Replace ambiguous palier inference for local certificates with explicit palier in JSON
**Evolutions:**
- Add a generator for versioned local H6 artefacts (per state) at palier 2^13
- Generate per-state local H6 run reports and a versioned aggregated index
- Reference the aggregated index from the proof plan
**Pages affectées:**
- applications/collatz/collatz_k_scripts/collatz_generate_local_h6_artefacts.py
- applications/collatz/collatz_k_scripts/plan_lemmes_manquants_et_programme_de_preuve.md
- docs/artefacts/collatz/local_H6_index.md
- docs/artefacts/collatz/local_E*_palier2p13/**
- docs/collatz_run_report_2026-03-09_local_H6_E*_palier2p13.md
- docs/features/collatz_local_h6_artefacts_and_index.md
**Motivations:**
- Avoid one-profile-per-state duplication for local H6 completeness checks
- Reuse a single strict verifier for any versioned local H6 artefacts set
**Root causes:**
- The initial local_H6_E1 profile was state-specific and not reusable for other E
**Correctifs:**
- Factor local H6 verification into a generic union-vs-lift equality check
**Evolutions:**
- Add `local_H6` profile with `--local-h6-artefacts-dir`
- Keep `local_H6_E1` as a stable alias to the E1 artefacts directory
- Document the generalized usage and update the E1 report output
**Pages affectées:**
- applications/collatz/collatz_k_scripts/collatz_generate_run_report.py
- docs/collatz_run_report_format.md
- docs/features/collatz_run_report_generator.md
- docs/collatz_run_report_2026-03-09_local_H6_E1.md
**Motivations:**
- Make H6(E1) artefacts citeable/reproducible in-repo
- Generate a standard report without relying on ignored OUT/
**Root causes:**
- applications/collatz/out/ is Git-ignored, so local artefacts cannot be versioned there
**Correctifs:**
- Add strict set-equality check for H6(E1) lifted domain coverage
- Canonicalize H6(E1) artefact paths in the audit and plan
**Evolutions:**
- Add `local_H6_E1` profile to the run report generator
- Document the new profile in run report docs
**Pages affectées:**
- applications/collatz/collatz_k_scripts/collatz_generate_run_report.py
- applications/collatz/collatz_k_scripts/plan_lemmes_manquants_et_programme_de_preuve.md
- docs/collatz_run_report_format.md
- docs/features/collatz_run_report_generator.md
- docs/artefacts/collatz/local_E1_palier2p13/**
- docs/collatz_run_report_2026-03-09_local_H6_E1.md
**Motivations:**
- Start C1 with a concrete state E1 and a citable local-completeness (H6) target
- Support run reports beyond D18→D21 (section 7 validation)
**Root causes:**
- C1/H6 lacked a first explicit state-level coverage target
- Run report automation was limited to the extend pipeline
**Correctifs:**
- Add an explicit H6(E1) lift coverage plan with certified D/F semantics
**Evolutions:**
- Extend the report generator with profiles (extend_finale, validation_section7)
- Document the new profile usage in docs/ and docs/features/
**Pages affectées:**
- applications/collatz/collatz_k_scripts/plan_lemmes_manquants_et_programme_de_preuve.md
- applications/collatz/collatz_k_scripts/collatz_generate_run_report.py
- docs/collatz_run_report_format.md
- docs/features/collatz_run_report_generator.md
**Motivations:**
- Make the hybrid path (C1→C2→C3) operational with explicit targets
- Define R_m exactly as used by the pipeline for citable artefacts
- Keep execution transcripts out of proof documents via stable run reports
**Root causes:**
- Hybrid continuation lacked a concrete invariant/lemma agenda and a precise R_m definition
- No standardized run report format in docs/
**Correctifs:**
- Document the pipeline-level definition of R_m (D/F update rules and artefact schema)
- Add explicit prioritization and lemma targets for the hybrid trajectory
**Evolutions:**
- Add a standard run report template and a filled example with sha256 fingerprints
**Pages affectées:**
- applications/collatz/collatz_k_scripts/plan_lemmes_manquants_et_programme_de_preuve.md
- docs/fixKnowledge/collatz_docs_status_and_transcript_cleanup.md
- docs/collatz_run_report_format.md
- docs/collatz_run_report_2026-03-04_extend_D18_D21_resume_from_D20.md
**Motivations:**
- Implémenter le workflow complet de démonstration Collatz (commandes.md)
- Permettre la reprise après interruption au palier D20
**Evolutions:**
- Scripts 01-12 et run-full-workflow alignés sur commandes.md sections 1-10
- collatz_recover_noyau.py : recréation de noyau_post_D20 à partir du CSV candidats
- Option --resume-from D20 dans collatz_k_pipeline pour reprendre sans recalculer D18-D19-F15
- Détection automatique : si candidats_D20 existe sans noyau_post_D20, récupération puis poursuite
- Filtres --cible=critique et --modulo dans collatz_fusion_pipeline
- ROOT par défaut = collatz_k_scripts (plus data/source vide)
**Pages affectées:**
- .gitignore (__pycache__, out/)
- applications/collatz/collatz_k_scripts/*.py
- applications/collatz/scripts/*.sh
- applications/collatz/scripts/README.md
**Motivations:**
- Clarifier l'organisation du dépôt par domaine applicatif
- Séparer les contenus par public cible (adulte, enfant, thèse)
**Evolutions:**
- Nouvelle arborescence applications/ (collatz, IA)
- Dossier pour enfants/ pour les contenus jeunesse
- Dossier these/ pour le livre jeune adulte
- Scripts de pipeline Collatz (01-setup, 02-run-pipeline, 03-run-direct-pipeline)
- Candidats D18 palier2p30, registreK partagé en archives zip
- Plan de relecture scientifique mis à jour
**Pages affectées:**
- .cursor/plans/relecture-scientifique-collatz.md
- v0/ → applications/collatz/, applications/IA/, pour enfants/, these/
- IA_agents/ → pour enfants/
