{"slug": "introducing-correctover-patronus-6-dimensional-verification-for-patronus-ai", "title": "Introducing correctover-patronus: 6-Dimensional Verification for Patronus AI", "summary": "Correctover released correctover-patronus, an adapter that integrates its 87 deterministic verification rules as native evaluators within the Patronus AI framework. The tool adds six verification dimensions—structure, schema, identity, integrity, latency, and cost—and generates a recomputable proof hash for each verdict, enabling verifiable output assurance for LLM applications.", "body_md": "LLM evaluation tools like Patronus AI excel at hallucination detection, toxicity checks, and semantic relevance. But they don't catch the *structural* failures:\n\nThese aren't hallucinations. They're verification failures.\n\n**correctover-patronus** is an adapter that runs Correctover's 87 deterministic verification rules as native Patronus evaluators. Every verdict comes with a recomputable proof hash — meaning you can verify the verifier.\n\n```\npip install correctover-patronus\n```\n\n| Dimension | What It Checks | Example |\n|---|---|---|\nStructure |\nOutput format validity | JSON parses correctly |\nSchema |\nField presence & types | Required fields exist |\nIdentity |\nSemantic relevance to input | Response addresses the question |\nIntegrity |\nForbidden pattern absence | No Tracebacks or error messages |\nLatency |\nResponse time budget | Under 30s threshold |\nCost |\nToken usage budget | Under 10k token limit |\n\n``` python\nfrom correctover_patronus import CorrectoverEvaluator, CorrectoverConfig\n\nconfig = CorrectoverConfig(\n    min_confidence=0.7,\n    latency_rules={\"max_ms\": 5000},\n    cost_rules={\"max_tokens\": 4000}\n)\n\nevaluator = CorrectoverEvaluator(config=config)\nresult = evaluator.evaluate(\n    task_input=\"Summarize this article...\",\n    task_output=\"The article discusses...\",\n    task_context={\"source\": \"article\", \"word_count\": 1500}\n)\n\nprint(f\"Overall: {result.score:.2f} ({'PASS' if result.pass_ else 'FAIL'})\")\nprint(f\"Proof hash: {result.metadata['proof_hash']}\")\nfor dim, info in result.metadata['dimensions'].items():\n    print(f\"  {dim}: {info['status']} (score={info['score']:.2f})\")\npython\nfrom correctover_patronus import correctover_structure, correctover_integrity\n\n# Check if output is valid JSON\nis_valid = correctover_structure(task_output='{\"key\": \"value\"}')\n\n# Check for error patterns\nis_clean = correctover_integrity(task_output=\"Result: 42\")\npython\nfrom correctover_patronus import correctover_full\n\n# Use in Patronus experiments for systematic benchmarking\nresults = patronus.evaluate(\n    evaluators=[correctover_full],\n    dataset=my_dataset,\n    experiment_name=\"correctover-benchmark\"\n)\n```\n\nEvery evaluation produces a `proof_hash`\n\nin the metadata. This hash covers:\n\nYou can re-run the same verification and get the same hash. No black boxes.\n\n```\n┌─────────────────┐     ┌──────────────────────┐\n│   Patronus AI   │────>│  correctover-patronus │\n│   Framework     │     │  (this adapter)       │\n└─────────────────┘     └──────────┬───────────┘\n                                   │\n                    ┌──────────────▼──────────────┐\n                    │     Correctover SDK         │\n                    │  (87 rules, 6 dimensions)   │\n                    │  P50 verification: 22us     │\n                    └──────────────┬──────────────┘\n                                   │\n                    ┌──────────────▼──────────────┐\n                    │   Verification Request      │\n                    │   -> Verdict + Proof Hash    │\n                    │   -> Metadata + Tags         │\n                    └─────────────────────────────┘\n```\n\n*Failover ≠ Correctover.*™\n\n*Correctover verifies. Patronus evaluates. Together: complete output assurance.", "url": "https://wpnews.pro/news/introducing-correctover-patronus-6-dimensional-verification-for-patronus-ai", "canonical_source": "https://dev.to/correctover/introducing-correctover-patronus-6-dimensional-verification-for-patronus-ai-cah", "published_at": "2026-07-01 06:27:17+00:00", "updated_at": "2026-07-01 06:48:34.482539+00:00", "lang": "en", "topics": ["large-language-models", "ai-tools", "developer-tools", "ai-products", "ai-infrastructure"], "entities": ["Correctover", "Patronus AI", "correctover-patronus"], "alternates": {"html": "https://wpnews.pro/news/introducing-correctover-patronus-6-dimensional-verification-for-patronus-ai", "markdown": "https://wpnews.pro/news/introducing-correctover-patronus-6-dimensional-verification-for-patronus-ai.md", "text": "https://wpnews.pro/news/introducing-correctover-patronus-6-dimensional-verification-for-patronus-ai.txt", "jsonld": "https://wpnews.pro/news/introducing-correctover-patronus-6-dimensional-verification-for-patronus-ai.jsonld"}}