StillHammer/seo-generator-server
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
34 Commits 2 Branches 0 Tags 1.4 MiB
74bf1b0f38
Commit Graph

4 Commits

Author SHA1 Message Date
StillHammer
0244521f5c feat(selective-smart-touch): Add intelligent analysis-driven enhancement system + validation spec 
## SelectiveSmartTouch (NEW)
- Architecture révolutionnaire: Analyse intelligente → Améliorations ciblées précises
- 5 modules: SmartAnalysisLayer, SmartTechnicalLayer, SmartStyleLayer, SmartReadabilityLayer, SmartTouchCore
- Système 10% segments: amélioration uniquement des segments les plus faibles (intensity-based)
- Détection contexte globale pour prompts adaptatifs multi-secteurs
- Intégration complète dans PipelineExecutor et PipelineDefinition

## Pipeline Validator Spec (NEW)
- Spécification complète système validation qualité par LLM
- 5 critères universels: Qualité, Verbosité, SEO, Répétitions, Naturalité
- Échantillonnage intelligent par filtrage balises (pas XML)
- Évaluation multi-versions avec justifications détaillées
- Coût estimé: ~$1/validation (260 appels LLM)

## Optimizations
- Réduction intensités fullEnhancement (technical 1.0→0.7, style 0.8→0.5)
- Ajout gardes-fous anti-familiarité excessive dans StyleLayer
- Sauvegarde étapes intermédiaires activée par défaut (pipeline-runner)

## Fixes
- Fix typo critique SmartTouchCore.js:110 (determineLayers ToApply → determineLayersToApply)
- Prompts généralisés multi-secteurs (e-commerce, SaaS, services, informatif)

🚀 Generated with Claude Code (https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
 2025-10-13 15:01:02 +08:00
StillHammer
64fb319e65 refactor: Synchronisation complète du codebase - Application de tous les patches 
Application systématique et méthodique de tous les patches historiques.

## ✅ FICHIERS SYNCHRONISÉS (19 fichiers)

### Core & Infrastructure:
- server.js (14 patches) - Lazy loading ModeManager, SIGINT hard kill, timing logs
- ModeManager.js (4 patches) - Instrumentation complète avec timing détaillé

### Pipeline System:
- PipelineDefinition.js (6 patches) - Source unique getLLMProvidersList()
- pipeline-builder.js (8 patches) - Standardisation LLM providers
- pipeline-runner.js (6 patches) - Affichage résultats structurés + debug console
- pipeline-builder.html (2 patches) - Fallback providers synchronisés
- pipeline-runner.html (3 patches) - UI améliorée résultats

### Enhancement Layers:
- TechnicalLayer.js (1 patch) - defaultLLM: 'gpt-4o-mini'
- StyleLayer.js (1 patch) - Type safety vocabulairePref
- PatternBreakingCore.js (1 patch) - Mapping modifications
- PatternBreakingLayers.js (1 patch) - LLM standardisé

### Validators & Tests:
- QualityMetrics.js (1 patch) - callLLM('gpt-4o-mini')
- PersonalityValidator.js (1 patch) - Provider gpt-4o-mini
- AntiDetectionValidator.js - Synchronisé

### Documentation:
- TODO.md (1 patch) - Section LiteLLM pour tracking coûts
- CLAUDE.md - Documentation à jour

### Tools:
- tools/analyze-skipped-exports.js (nouveau)
- tools/apply-claude-exports.js (nouveau)
- tools/apply-claude-exports-fuzzy.js (nouveau)

## 🎯 Changements principaux:
- ✅ Standardisation LLM providers (openai → gpt-4o-mini, claude → claude-sonnet-4-5)
- ✅ Lazy loading optimisé (ModeManager chargé à la demande)
- ✅ SIGINT immediate exit (pas de graceful shutdown)
- ✅ Type safety renforcé (conversions string explicites)
- ✅ Instrumentation timing complète
- ✅ Debug logging amélioré (console.log résultats pipeline)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
 2025-10-12 20:36:17 +08:00
StillHammer
3751ab047b feat(keywords): Add hierarchical context to missing keywords prompt and fix LLM response format 
This commit improves keyword generation by providing hierarchical context for each element and fixing the LLM response format parsing.

Changes:
1. lib/MissingKeywords.js:
   - Add buildHierarchicalContext() to generate compact contextual info for each element
   - Display hierarchy in prompt (e.g., "H2 existants: 'Titre1', 'Titre2'")
   - For Txt elements: show associated MC keyword + parent title
   - For FAQ elements: count existing FAQs
   - Fix LLM response format by providing 3 concrete examples from actual list
   - Add explicit warning to use exact tag names [Titre_H2_3], [Txt_H2_6]
   - Improve getElementContext() to better retrieve hierarchical elements

2. lib/selective-enhancement/SelectiveUtils.js:
   - Fix createTypedPrompt() to use specific keyword from resolvedContent
   - Remove fallback to csvData.mc0 (log error if no specific keyword)

3. lib/pipeline/PipelineExecutor.js:
   - Integrate generateMissingSheetVariables() as "Étape 0" before extraction

Prompt format now:
  1. [Titre_H2_3] (titre) — H2 existants: "Titre1", "Titre2"
  2. [Txt_H2_6] (texte) — MC: "Plaque dibond" | Parent: "Guide dibond"
  3. [Faq_q_1] (question) — 3 FAQ existantes

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
 2025-10-12 14:51:01 +08:00
StillHammer
471058f731 Add flexible pipeline system with per-module LLM configuration 
- New modular pipeline architecture allowing custom workflow combinations
- Per-step LLM provider configuration (Claude, OpenAI, Gemini, Deepseek, Moonshot, Mistral)
- Visual pipeline builder and runner interfaces with drag-and-drop
- 10 predefined pipeline templates (minimal-test to originality-bypass)
- Pipeline CRUD operations via ConfigManager and REST API
- Fix variable resolution in instructions (HTML tags were breaking {{variables}})
- Fix hardcoded LLM providers in AdversarialCore
- Add TESTS_LLM_PROVIDER.md documentation with validation results
- Update dashboard to disable legacy config editor

API Endpoints:
- POST /api/pipeline/save, execute, validate, estimate
- GET /api/pipeline/list, modules, templates

Backward compatible with legacy modular workflow system.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
 2025-10-09 14:01:52 +08:00