Protokoll #17790

{
    "event": "PreToolUse",
    "tool_name": "Read",
    "tool_input": {
        "file_path": "\/var\/www\/scripts\/pipeline\/analyzers\/entity_extractor.py",
        "offset": 105,
        "limit": 80
    }
}

{
    "tool_response": {
        "type": "text",
        "file": {
            "filePath": "\/var\/www\/scripts\/pipeline\/analyzers\/entity_extractor.py",
            "content": "\n            # Strict validation: entity must appear EXACTLY in source text\n            if not _validate_entity_in_text(item, source_text):\n                continue  # Skip hallucinations\n\n            entities.append({\n                \"name\": item,\n                \"type\": entity_type,\n                \"description\": None,\n            })\n\n    return entities\n\n\ndef extract_entities_ollama(text: str, model: str = \"gemma3:27b-it-qat\") -> list[dict]:\n    \"\"\"Extract entities using Ollama.\n\n    Supports two response formats:\n    1. New format: {\"persons\":[], \"roles\":[], \"theories\":[], ...}\n    2. Legacy format: {\"entities\": [{\"name\": \"...\", \"type\": \"...\"}]}\n    \"\"\"\n    prompt_data = db.get_prompt_by_use_case(\"entity_extraction\")\n    prompt_template = prompt_data[\"content\"] if prompt_data else None\n\n    if not prompt_template:\n        db.log(\"WARNING\", \"entity_extraction prompt not found in DB, using fallback\")\n        prompt_template = \"\"\"Analysiere den Text und extrahiere wichtige Entitäten.\nKategorisiere als: PERSON, ORGANIZATION, CONCEPT, LOCATION\nAntworte NUR im JSON-Format:\n{\"entities\": [{\"name\": \"...\", \"type\": \"...\", \"description\": \"...\"}]}\n\nText:\n{text}\"\"\"\n\n    # Support both {text} and {{TEXT}} placeholders\n    prompt = prompt_template.replace(\"{text}\", text[:3000]).replace(\"{{TEXT}}\", text[:3000])\n\n    try:\n        start_time = time.time()\n        response = requests.post(\n            f\"{OLLAMA_HOST}\/api\/generate\",\n            json={\"model\": model, \"prompt\": prompt, \"stream\": False, \"format\": \"json\"},\n            timeout=120,\n        )\n        response.raise_for_status()\n        data = response.json()\n        duration_ms = int((time.time() - start_time) * 1000)\n\n        response_text = data.get(\"response\", \"{}\")\n\n        protokoll.log_llm_call(\n            request=f\"[entity_extraction] {prompt[:500]}...\",\n            response=response_text[:2000],\n            model_name=f\"ollama:{model}\",\n            tokens_input=data.get(\"prompt_eval_count\", 0),\n            tokens_output=data.get(\"eval_count\", 0),\n            duration_ms=duration_ms,\n            status=\"completed\",\n        )\n\n        try:\n            result = json.loads(response_text)\n            return _normalize_entity_response(result, text)\n        except json.JSONDecodeError:\n            db.log(\"WARNING\", \"Failed to parse entity JSON from Ollama\")\n            return []\n    except Exception as e:\n        db.log(\"ERROR\", f\"Ollama entity extraction failed: {e}\")\n        protokoll.log_llm_call(\n            request=f\"[entity_extraction] {prompt[:500]}...\",\n            model_name=f\"ollama:{model}\",\n            status=\"error\",\n            error_message=str(e),\n        )\n        return []\n\n\ndef extract_entities_anthropic(text: str, client) -> list[dict]:\n    \"\"\"Extract entities using Anthropic Claude.\"\"\"\n    prompt_data = db.get_prompt_by_use_case(\"entity_extraction\")",
            "numLines": 80,
            "startLine": 105,
            "totalLines": 354
        }
    }
}