{"slug": "jevotron-multiple-jev-integrations-from-the-command-line", "title": "Jevotron: Multiple Jev integrations from the command line", "summary": "Jevotron launched a command-line tool that scans CSV, YAML, JSON, TOML, text, and OBO files for field-level anomalies using the TypeSafe API, caching each successful assessment in SQLite so unchanged entries are not reassessed. In a pilot on 24 public agent traces, Jevotron matched 130 of 163 step-quality labels (79.8%), with 89.7% precision and 70.3% recall on harmful-step detection. The tool requires a TYPESAFE_API_KEY for scans, while its preview command runs without a key.", "body_md": "# One-shot anomaly detection on the command line[¶](#one-shot-anomaly-detection-on-the-command-line)\n\nGive jevotron a file and a little guidance. Get field-level anomaly scores, a focused review queue, and a cache that makes the next run cheaper.\n\n[Run your first scan](quickstart/)\n[Explore the commands](reference/cli/)\n[Install agent skills](guides/agents/)\n\n``` bash\n$ jevotron scan airports.csv --guidance \"Check airport locations.\"\n```\n\nFirst, set your API key\n\nGet a key from the [TypeSafe dashboard](https://console.typesafe.ai/) and set it in your shell:\n\n```\nexport TYPESAFE_API_KEY=\"your-api-key\"\n```\n\nAlready set? You're ready to scan. `preview` needs no key.\n[API key setup →](quickstart/#set-your-api-key) · [See the actual airport results →](examples/airports/#actual-output)\n\n01 / PREVIEW\n\n### See what goes in[¶](#see-what-goes-in)\n\nInspect chunks, field paths, and exact model requests before making an API call.\n\n02 / SCAN\n\n### Assess every entry[¶](#assess-every-entry)\n\nJev scores selected fields together. Each entry gets the same guidance and optional examples.\n\n03 / REVIEW\n\n### Start with the warnings[¶](#start-with-the-warnings)\n\nSort suspicious entries, export CSV, or pipe JSONL into your existing shell workflow.\n\n## Small setup. Useful defaults.[¶](#small-setup-useful-defaults)\n\n**CSV, YAML, JSON, TOML, text, OBO, and more** work out of the box, including\ngzip files. The [format reference](reference/formats/) covers defaults and\nformat-specific options. Select fields with\n`--field`, add a sentence with `--guidance`, and run. Longer instructions can\ncome from `--guidance-file`. A local Python config is\navailable when a project needs custom parsing or reusable settings.\n\n**Unchanged input reuses its assessment.** SQLite saves each successful result as\nit arrives. Reorder a file, change a reporting threshold, or resume a failed run\nwithout reassessing unchanged entries.\n\n**A review aid with visible evidence.** Reports retain each field's probabilities,\nthe entry score, source location, and assessment date. The warning score is the\nhighest field anomaly probability; you choose the threshold.\n\n## Try a complete example[¶](#try-a-complete-example)\n\n| Example | What you'll do | \n|---|---|\n| [Airports / CSV](examples/airports/) | Find two injected country errors in public data, then compare versions. | \n| [Inventory / YAML](examples/inventory/) | Apply written rules and a chosen exemplar to stock records. | \n| [Measurement units / OBO](examples/obo/) | Score definitions and repeated synonyms within independent stanzas. | \n| [Agent traces / JSONL](examples/agent-traces/) | Classify public agent traces and individual steps, then compare with published labels. | \n\n### Agent traces: a measured pilot[¶](#agent-traces-a-measured-pilot)\n\nOn a small, length-filtered sample of **24 public traces**, jt matched **130 of\n163 step-quality labels (79.8%)**. Harmful-step precision was **89.7%**, with\n**70.3% recall**. The example uses original messages and tool definitions, with\nhuman labels withheld from the model.\n\n[Try the trace example →](examples/agent-traces/) ·\n[Read the full analysis and limitations →](analysis/agent-traces/)", "url": "https://wpnews.pro/news/jevotron-multiple-jev-integrations-from-the-command-line", "canonical_source": "https://cmungall.github.io/jevotron/", "published_at": "2026-10-01 02:12:38+00:00", "updated_at": "2026-10-01 02:18:21.611282+00:00", "lang": "en", "topics": ["ai-tools", "ai-agents", "developer-tools", "mlops"], "entities": ["Jevotron", "TypeSafe", "SQLite", "TYPESAFE_API_KEY"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/jevotron-multiple-jev-integrations-from-the-command-line", "markdown": "https://wpnews.pro/news/jevotron-multiple-jev-integrations-from-the-command-line.md", "text": "https://wpnews.pro/news/jevotron-multiple-jev-integrations-from-the-command-line.txt", "jsonld": "https://wpnews.pro/news/jevotron-multiple-jev-integrations-from-the-command-line.jsonld"}}