{
"tool_response": {
"filePath": "\/var\/www\/scripts\/pipeline\/pipeline.py",
"oldString": "def run_full_pipeline(run_id=None, pipeline_id=None):\n \"\"\"Run complete pipeline: scan → process → embed.\"\"\"\n progress = PipelineProgress(run_id) if run_id else None\n\n print(\"=\" * 50)\n print(\"KI-System Pipeline - Full Run\")\n if run_id:\n print(f\"Run ID: {run_id}, Pipeline ID: {pipeline_id}\")\n print(\"=\" * 50)",
"newString": "def run_full_pipeline(run_id=None, pipeline_id=None):\n \"\"\"Run complete pipeline: scan → process → embed.\n\n Pipeline selection:\n - pipeline_id=5: Scientific Pipeline (semantics BEFORE embedding)\n - pipeline_id=1-4 or None: Legacy Pipeline (semantics AFTER embedding)\n \"\"\"\n progress = PipelineProgress(run_id) if run_id else None\n\n # Determine which processing function to use\n use_v5 = pipeline_id == 5\n pipeline_name = \"Wissenschaftliche Pipeline v1\" if use_v5 else \"Standard Pipeline\"\n\n print(\"=\" * 50)\n print(f\"KI-System Pipeline - {pipeline_name}\")\n if run_id:\n print(f\"Run ID: {run_id}, Pipeline ID: {pipeline_id}\")\n if use_v5:\n print(\"Mode: Semantik VOR Embedding (wissenschaftlich korrekt)\")\n print(\"=\" * 50)",
"originalFile": "#!\/usr\/bin\/env python3\n\"\"\"\nKI-System Document Pipeline\nMain orchestration script for document processing.\n\nUsage:\n python pipeline.py scan # Scan for new documents\n python pipeline.py process # Process queued documents\n python pipeline.py embed # Embed pending chunks\n python pipeline.py semantic <id> # Run semantic analysis on document\n python pipeline.py semantic-queue # Process semantic queue\n python pipeline.py enrich-entities # Enrich entity descriptions via Ollama\n python pipeline.py enrich-entities 50 # Limit to 50 entities\n python pipeline.py all # Full pipeline run\n python pipeline.py all --pipeline-id=1 --run-id=5 # With tracking\n python pipeline.py file <path> # Process single file\n python pipeline.py status # Show pipeline status\n\"\"\"\n\nimport argparse\nimport os\nimport time\nfrom pathlib import Path\n\nfrom config import (\n MAX_RETRIES,\n RETRY_BACKOFF_BASE,\n SEMANTIC_AUTO_QUEUE,\n SEMANTIC_SYNC,\n SEMANTIC_USE_ANTHROPIC,\n)\nfrom constants import DEFAULT_LIMIT\nfrom db import PipelineProgress, db\nfrom detect import queue_files, scan_directory\nfrom step_embed import EmbeddingStep\nfrom step_entity_enrich import EntityEnrichStep\nfrom step_extract import ExtractionStep\nfrom step_load import LoadStep\nfrom step_semantic import SemanticStep\nfrom step_semantic_extended import (\n DuplicateCheckStep,\n KnowledgeSemanticAnalyzeStep,\n KnowledgeSemanticStoreStep,\n TextSemanticAnalyzeStep,\n TextSemanticStoreStep,\n)\nfrom step_transform import TransformationStep\n\n\ndef process_file(file_path, progress=None):\n \"\"\"Process a single file through the pipeline.\"\"\"\n file_name = Path(file_path).name\n\n if progress:\n progress.update_document(file_name)\n\n # Initialize pipeline steps\n extract_step = ExtractionStep(db, progress)\n load_step = LoadStep(db, progress)\n transform_step = TransformationStep(db, progress)\n embed_step = EmbeddingStep(db, progress)\n\n # Check if cancelled before starting\n if progress and progress.is_cancelled():\n return \"cancelled\", 0, 0\n\n # Step 1: Extract\n extract_result = extract_step.execute(file_path)\n if not extract_result[\"success\"]:\n if extract_result.get(\"error\") == \"cancelled\":\n return \"cancelled\", 0, 0\n return False, 0, 0\n\n extraction = extract_result[\"extraction\"]\n file_info = extract_result[\"file_info\"]\n total_pages = extract_result.get(\"total_pages\", 0)\n\n # Check if cancelled after extraction\n if progress and progress.is_cancelled():\n return \"cancelled\", 0, 0\n\n # Step 2: Load document\n doc_id = load_step.create_document(file_info)\n\n # Step 3: Store pages (PDFs and multi-page documents)\n page_map = load_step.store_pages(doc_id, extraction)\n\n # Step 4: Vision analysis (PDFs only)\n if file_info[\"type\"] == \".pdf\":\n transform_step.execute_vision(doc_id, file_path, file_info[\"type\"])\n\n # Check if cancelled after vision\n if progress and progress.is_cancelled():\n return \"cancelled\", 0, 0\n\n # Step 5: Chunking\n chunks = transform_step.execute_chunking(extraction, total_pages)\n\n # Step 6: Store chunks with page references\n chunks = load_step.store_chunks(doc_id, chunks, page_map)\n\n # Check if cancelled after chunking\n if progress and progress.is_cancelled():\n return \"cancelled\", len(chunks), 0\n\n # Step 7: Enrichment (PDFs only)\n if file_info[\"type\"] == \".pdf\":\n transform_step.execute_enrichment(doc_id, file_info[\"type\"])\n\n # Check if cancelled after enrichment\n if progress and progress.is_cancelled():\n return \"cancelled\", len(chunks), 0\n\n # Step 8: Embeddings (Layer 3 - Document becomes searchable)\n embedded = embed_step.execute(chunks, doc_id, file_name, file_path)\n\n # Document is now searchable - update status to \"embedded\"\n load_step.update_document_status(doc_id, \"embedded\")\n\n if progress:\n progress.add_log(f\"Layer 3 fertig: {file_name} ist jetzt suchbar\")\n\n # Check if cancelled after embedding\n if progress and progress.is_cancelled():\n return \"cancelled\", len(chunks), embedded\n\n # Step 9: Semantic analysis (Layer 4 - Optional\/Async)\n semantic_step = SemanticStep(db, progress)\n full_text = extract_step.get_full_text_from_extraction(extraction)\n\n if SEMANTIC_SYNC:\n # Run semantic analysis synchronously\n try:\n semantic_step.execute(doc_id, full_text, use_anthropic=SEMANTIC_USE_ANTHROPIC)\n # Update to done only after semantic completes\n load_step.update_document_status(doc_id, \"done\")\n except Exception as e:\n # Semantic failed but document is still searchable\n db.log(\"WARNING\", f\"Semantic analysis failed for {file_name}: {e}\")\n if progress:\n progress.add_log(f\"Semantik-Fehler (Dokument bleibt suchbar): {str(e)[:50]}\")\n elif SEMANTIC_AUTO_QUEUE:\n # Queue for async processing\n semantic_step.queue(doc_id, priority=5)\n load_step.update_document_status(doc_id, \"done\")\n if progress:\n progress.add_log(f\"Semantik in Queue: {file_name}\")\n else:\n # No semantic analysis\n load_step.update_document_status(doc_id, \"done\")\n\n if progress:\n progress.add_log(f\"Fertig: {file_name}\")\n\n return True, len(chunks), embedded\n\n\ndef process_file_v5(file_path, progress=None):\n \"\"\"Process a single file through Pipeline #5 (Scientific Pipeline).\n\n Key difference from process_file():\n - Semantic analysis happens BEFORE embedding (scientifically correct)\n - Uses extended semantic steps for text and knowledge semantics\n \"\"\"\n file_name = Path(file_path).name\n\n if progress:\n progress.update_document(file_name)\n\n # Initialize pipeline steps\n extract_step = ExtractionStep(db, progress)\n load_step = LoadStep(db, progress)\n transform_step = TransformationStep(db, progress)\n embed_step = EmbeddingStep(db, progress)\n text_semantic_analyze = TextSemanticAnalyzeStep(db, progress)\n text_semantic_store = TextSemanticStoreStep(db, progress)\n knowledge_semantic_analyze = KnowledgeSemanticAnalyzeStep(db, progress)\n knowledge_semantic_store = KnowledgeSemanticStoreStep(db, progress)\n duplicate_check = DuplicateCheckStep(db, progress)\n\n # Check if cancelled before starting\n if progress and progress.is_cancelled():\n return \"cancelled\", 0, 0\n\n # Phase 1: Existenz - Extract\n extract_result = extract_step.execute(file_path)\n if not extract_result[\"success\"]:\n if extract_result.get(\"error\") == \"cancelled\":\n return \"cancelled\", 0, 0\n return False, 0, 0\n\n extraction = extract_result[\"extraction\"]\n file_info = extract_result[\"file_info\"]\n total_pages = extract_result.get(\"total_pages\", 0)\n content_hash = file_info.get(\"hash\", \"\")\n\n # Check if cancelled after extraction\n if progress and progress.is_cancelled():\n return \"cancelled\", 0, 0\n\n # Phase 1: Existenz - Load document\n doc_id = load_step.create_document(file_info)\n\n # Phase 1: Existenz - Duplicate check\n dup_result = duplicate_check.execute(doc_id, content_hash)\n if dup_result[\"status\"] == \"abort\":\n load_step.update_document_status(doc_id, \"duplicate\")\n if progress:\n progress.add_log(f\"Duplikat: {file_name} = Doc #{dup_result['duplicate_id']}\")\n return True, 0, 0 # Not an error, just skip\n\n # Phase 2: Text - Store pages\n page_map = load_step.store_pages(doc_id, extraction)\n\n # Phase 2: Text - Vision analysis (PDFs only)\n if file_info[\"type\"] == \".pdf\":\n transform_step.execute_vision(doc_id, file_path, file_info[\"type\"])\n if progress and progress.is_cancelled():\n return \"cancelled\", 0, 0\n\n # Phase 3: Struktur - Chunking\n chunks = transform_step.execute_chunking(extraction, total_pages)\n\n # Phase 3: Struktur - Store chunks with page references\n chunks = load_step.store_chunks(doc_id, chunks, page_map)\n\n if progress and progress.is_cancelled():\n return \"cancelled\", len(chunks), 0\n\n # Phase 3: Struktur - Enrichment (PDFs only)\n if file_info[\"type\"] == \".pdf\":\n transform_step.execute_enrichment(doc_id, file_info[\"type\"])\n if progress and progress.is_cancelled():\n return \"cancelled\", len(chunks), 0\n\n # Phase 4: Textsemantik - Analyze chunks\n if progress:\n progress.add_log(\"Phase 4: Textsemantik...\")\n\n # Prepare chunks for analysis\n chunk_data = [{\"id\": c[\"id\"], \"content\": c[\"content\"]} for c in chunks]\n analyzed_chunks = text_semantic_analyze.execute(chunk_data, {\"model\": \"mistral\"})\n\n # Store text semantics\n text_semantic_store.execute(analyzed_chunks, {})\n\n if progress and progress.is_cancelled():\n return \"cancelled\", len(chunks), 0\n\n # Phase 5-6: Entity + Wissenssemantik\n if progress:\n progress.add_log(\"Phase 5-6: Entity-Extraktion + Wissenssemantik...\")\n\n # Run standard semantic analysis (entities, relations, taxonomy)\n semantic_step = SemanticStep(db, progress)\n full_text = extract_step.get_full_text_from_extraction(extraction)\n\n try:\n semantic_step.execute(doc_id, full_text, use_anthropic=SEMANTIC_USE_ANTHROPIC)\n except Exception as e:\n db.log(\"WARNING\", f\"Semantic analysis failed for {file_name}: {e}\")\n if progress:\n progress.add_log(f\"Semantik-Warnung: {str(e)[:50]}\")\n\n # Load entities for knowledge semantics\n cursor = db.execute(\n \"\"\"SELECT e.id, e.name, e.type, c.content as context\n FROM entities e\n JOIN chunk_entities ce ON e.id = ce.entity_id\n JOIN chunks c ON ce.chunk_id = c.id\n WHERE c.document_id = %s\n GROUP BY e.id\"\"\",\n (doc_id,),\n )\n entities = cursor.fetchall()\n cursor.close()\n\n if entities:\n # Convert... [TRUNCATED-61ec1ca2625e82b5]",
"structuredPatch": [
{
"oldStart": 351,
"oldLines": 13,
"newStart": 351,
"newLines": 24,
"lines": [
" ",
" ",
" def run_full_pipeline(run_id=None, pipeline_id=None):",
"- \"\"\"Run complete pipeline: scan → process → embed.\"\"\"",
"+ \"\"\"Run complete pipeline: scan → process → embed.",
"+",
"+ Pipeline selection:",
"+ - pipeline_id=5: Scientific Pipeline (semantics BEFORE embedding)",
"+ - pipeline_id=1-4 or None: Legacy Pipeline (semantics AFTER embedding)",
"+ \"\"\"",
" progress = PipelineProgress(run_id) if run_id else None",
" ",
"+ # Determine which processing function to use",
"+ use_v5 = pipeline_id == 5",
"+ pipeline_name = \"Wissenschaftliche Pipeline v1\" if use_v5 else \"Standard Pipeline\"",
"+",
" print(\"=\" * 50)",
"- print(\"KI-System Pipeline - Full Run\")",
"+ print(f\"KI-System Pipeline - {pipeline_name}\")",
" if run_id:",
" print(f\"Run ID: {run_id}, Pipeline ID: {pipeline_id}\")",
"+ if use_v5:",
"+ print(\"Mode: Semantik VOR Embedding (wissenschaftlich korrekt)\")",
" print(\"=\" * 50)",
" ",
" try:"
]
}
],
"userModified": false,
"replaceAll": false
}
}