11 Commits

Author	SHA1	Message	Date
Lars	1f8791f4dd	feat: Phase 2 - Normalisierung + Workflow Executor ... Backend: - normalization_engine.py (200 Zeilen): Synonym-Mapping, 5 Statuswerte * normalize_decision_signal(): Kaskade (exact → case → synonym → invalid) * apply_synonym_mapping(): DB-basierte Synonyme (case-insensitive) * normalize_all_signals(): Batch-Processing gegen Katalog * load_question_catalog(): Lädt normalization_rules aus DB - workflow_executor.py (440 Zeilen): Sequenzielle Workflow-Ausführung * execute_workflow(): Traversiert DAG in topologischer Reihenfolge * execute_node(): Führt analysis nodes aus (start/end = no-op) * aggregate_results(): Kombiniert analysis_core + normalized_signals * save_execution_state(): Persistiert in workflow_executions - workflow_models.py: Erweitert um Phase 2 Models * SignalStatus Enum (valid, normalized, unclear, invalid, not_decidable) * NormalizedSignal (question_type, raw_value, normalized_value, status) * NodeExecutionState (node_id, status, analysis_core, normalized_signals) * ExecutionResult (execution_id, workflow_id, status, node_states, aggregated_result) - workflow_engine.py: Neue Funktion get_execution_order() * Flattened topological sort für sequenzielle Execution * Phase 7: Wird zu levels (parallele Execution) - prompt_executor.py: execute_workflow_prompt() Implementierung * Ruft workflow_executor.execute_workflow() auf * Konvertiert ExecutionResult zu API-Response - routers/workflows.py (230 Zeilen): Workflow Execution API * POST /api/workflows/{id}/execute (mit enable_debug) * GET /api/workflows/executions/{id} (lädt gespeicherten State) * GET /api/workflows (listet alle aktiven Workflows) * GET /api/workflows/{id} (lädt einzelnen Workflow mit Graph) - main.py: Router-Registrierung (workflows.router) Tests: - test_phase2_normalization.py (17 Tests): Alle Normalisierungs-Szenarien * Exact match, case-insensitive, synonym mapping, invalid, whitespace * Batch-Normalisierung, not_in_catalog, mixed validity - test_phase2_workflow_executor.py (10 Tests): Executor + Aggregation * aggregate_results mit verschiedenen Konstellationen * execute_node für start/end/analysis/unknown * Integration mit question_augmenter + result_container_parser Alle 27 Unit-Tests bestanden. version: 0.9k (backend) module: workflow 0.3.0 Konzept: .claude/task/Workflow_engine_prompting_engine/anforderungsanalyse_umsetzungsplan.md (Phase 2)	2026-04-03 21:20:23 +02:00
Lars	ca562b7130	feat: Phase 1 - Fragenergänzung + Strukturierter Container ... Backend: - question_augmenter.py (290 Zeilen): Hybrid-Modell für Fragenergänzungen * merge_question_augmentations(): Knotengebundene Fragen überschreiben Prompt-Defaults * augment_prompt_with_questions(): Markdown-formatierte Fragenergänzung * parse_question_augmentations_from_jsonb(): JSONB → QuestionAugmentation[] - result_container_parser.py (250 Zeilen): Markdown-Sektionen-Parsing * parse_result_container(): Extrahiert Analysekern, Entscheidungsanteil, Begründungsanker * validate_decision_signal(): Normalisierung gegen answer_spectrum * Fallback-Parsing bei unstrukturierten Antworten - routers/workflow_questions.py (236 Zeilen): CRUD für workflow_question_catalog * GET /api/workflow/questions (mit active_only Filter) * POST/PUT/DELETE (Admin only, Soft Delete) - prompt_executor.py: Integration in execute_base_prompt() * Fragenergänzung vor LLM-Call (wenn node_questions oder catalog vorhanden) * Result-Container-Parsing nach LLM-Response - main.py: Router-Registrierung (workflow_questions) Tests: - test_phase1_question_augmenter.py (8 Tests): Hybrid-Modell, Formatierung, JSONB-Parsing - test_phase1_result_container_parser.py (17 Tests): Sektion-Extraktion, Decision-Parsing, Validierung Alle 25 Unit-Tests bestanden. version: 0.9j (backend) module: workflow 0.2.0 Konzept: .claude/task/Workflow_engine_prompting_engine/konzept_workflow_engine_konsolidated.md (Phase 1)	2026-04-03 18:02:25 +02:00
Lars	b5be6e21a5	feat: Phase 0 - Workflow Engine Foundation ... Backend: - DB-Migration 034: workflow_definitions, workflow_question_catalog, workflow_executions - ai_prompts.question_augmentations JSONB-Spalte (Hybridmodell: Prompt-Defaults) - 6 Grundtypen Fragenergänzungen mit Normalisierungsregeln (Seed-Daten) - Pydantic-Modelle (16 Models, 11 Enums) in workflow_models.py - Workflow-Engine: Graph-Parsing, Topologische Sortierung, DAG-Validierung - Dispatcher-Erweiterung type='workflow' (Stub für Phase 1-3) - Adjacency Lists, Erreichbarkeits-Prüfungen, Zyklen-Erkennung Testing: - 22 Unit-Tests (alle bestanden): Graph-Parsing, Validierung, Topologische Sortierung - Fixtures: simple_valid_graph, parallel_graph, branching_graph Version: - APP_VERSION 0.9i - DB_SCHEMA_VERSION 20260403 - Module: workflow 0.1.0 Anforderungsanalyse: .claude/task/Workflow_engine_prompting_engine/anforderungsanalyse_umsetzungsplan.md Konzept-Basis: .claude/task/Workflow_engine_prompting_engine/konzept_workflow_engine_konsolidated.md Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>	2026-04-03 16:55:51 +02:00
Lars	6e651b5bb5	fix: include stage outputs in debug info for value table ... - stage_debug now includes 'output' dict with all stage outputs - Fixes empty values for stage_X_outputkey in expert mode - Stage outputs are the actual AI responses passed to next stage	2026-03-26 14:33:00 +01:00
Lars	4a2bebe249	fix: value table metadata + |d modifier + cursor insertion (Issues #47, #48) ... BUG: Wertetabelle wurde nicht angezeigt FIX: enable_debug=true wenn save=true (für metadata collection) - metadata wird nur gespeichert wenn debug aktiv - jetzt: debug or save → metadata immer verfügbar BUG: {{placeholder|d}} Modifier funktionierte nicht ROOT CAUSE: catalog wurde bei Exception nicht zu variables hinzugefügt FIX: - variables['_catalog'] = catalog (auch wenn None) - Warning-Log wenn catalog nicht geladen werden kann - Debug warning wenn |d ohne catalog verwendet BUG: Platzhalter in Pipeline-Stages am Ende statt an Cursor FIX: - stageTemplateRefs Map für alle Stage-Textareas - onClick + onKeyUp tracking für Cursor-Position - Insert at cursor: template.slice(0, pos) + placeholder + template.slice(pos) - Focus + Cursor restore nach Insert TECHNICAL: - prompt_executor.py: Besseres Exception Handling für catalog - UnifiedPromptModal.jsx: Refs für alle Template-Felder - prompts.py: enable_debug=debug or save version: 9.6.1 (bugfix) module: prompts 2.1.1	2026-03-26 12:04:20 +01:00
Lars	c0a50dedcd	feat: value table + {{placeholder|d}} modifier (Issue #47) ... FEATURE #47: Wertetabelle nach KI-Analysen - Migration 021: metadata JSONB column in ai_insights - Backend sammelt resolved placeholders mit descriptions beim Speichern - Frontend: Collapsible value table in InsightCard - Zeigt: Platzhalter | Wert | Beschreibung - Sortiert tabellarisch - Funktioniert für base + pipeline prompts FEATURE #48: {{placeholder|d}} Modifier - Syntax: {{weight_aktuell|d}} → "85.2 kg (Aktuelles Gewicht in kg)" - resolve_placeholders() erkennt |d modifier - Hängt description aus catalog an Wert - Fein-granulare Kontrolle pro Platzhalter (nicht global) - Optional: nur wo sinnvoll einsetzen TECHNICAL: - prompt_executor.py: catalog parameter durchgereicht - execute_prompt_with_data() lädt catalog via get_placeholder_catalog() - Catalog als _catalog in variables übergeben, in execute_prompt() extrahiert - Base + Pipeline Prompts unterstützen |d modifier EXAMPLE: Template: "Gewicht: {{weight_aktuell|d}}, Alter: {{age}}" Output: "Gewicht: 85.2 kg (Aktuelles Gewicht in kg), Alter: 55" version: 9.6.0 (feature) module: prompts 2.1.0, insights 1.4.0	2026-03-26 11:52:26 +01:00
Lars	ba92d66880	fix: remove {{ }} from placeholder keys before resolution ... Placeholder resolver returns keys with {{ }} wrappers, but resolve_placeholders expects clean keys. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>	2026-03-26 08:17:22 +01:00
Lars	afc70b5a95	fix: integrate placeholder resolver + JSON unwrapping (Issue #28) ... - Backend: integrate get_placeholder_example_values in execute_prompt_with_data - Backend: now provides BOTH raw data AND processed placeholders - Backend: unwrap Markdown-wrapped JSON (```json ... ```) - Fixes old-style prompts that expect name, weight_trend, caliper_summary Resolves unresolved placeholders issue. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>	2026-03-26 08:14:41 +01:00
Lars	84dad07e15	fix: show debug info on errors + prompt export function ... - Frontend: debug viewer now shows even when test fails - Frontend: export button to download complete prompt config as JSON - Backend: attach debug info to JSON validation errors - Backend: include raw output and length in error details Users can now debug failed prompts and export configs for analysis. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>	2026-03-26 08:07:34 +01:00
Lars	7f2ba4fbad	feat: debug system for prompt execution (Issue #28) ... - Backend: debug mode in prompt_executor with placeholder tracking - Backend: show resolved/unresolved placeholders, final prompts, AI responses - Frontend: test button in UnifiedPromptModal for saved prompts - Frontend: debug output viewer with JSON preview - Frontend: wider placeholder example fields in PlaceholderPicker Resolves pipeline execution debugging issues. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>	2026-03-26 08:01:33 +01:00
Lars	7be7266477	feat: unified prompt executor - Phase 2 complete (Issue #28) ... Backend: - prompt_executor.py: Universal executor for base + pipeline prompts - Dynamic placeholder resolution - JSON output validation - Multi-stage parallel execution (sequential impl) - Reference and inline prompt support - Data loading per module (körper, ernährung, training, schlaf, vitalwerte) Endpoints: - POST /api/prompts/execute - Execute unified prompts - POST /api/prompts/unified - Create unified prompts - PUT /api/prompts/unified/{id} - Update unified prompts Frontend: - api.js: executeUnifiedPrompt, createUnifiedPrompt, updateUnifiedPrompt Next: Phase 3 - Frontend UI consolidation	2026-03-25 14:52:24 +01:00