mc-jev-cli

Terminal and browser client for Ollama's /v1/systemone structured-inference endpoint.

JEV is named after economist William Stanley Jevons and is a System One Model — a new class of AI designed for automation, not conversation. It produces type-safe structured values (not text) in 70–500 ms, with hallucinations mathematically prevented. Learn more: typesafe.ai/blog/introducing-system-one-models-and-jev

Screenshot

What it does

systemone is an Ollama endpoint that answers structured questions about a piece of text in a single call.
Each question is typed:

| Type | What it returns | |------|----------------| | choice | Picks the best option from a named set, with probabilities for each | | noul | Returns a 0–1 score (null-or-likely, i.e. boolean strength) | | score | Returns a continuous 0–N score mapped over a legend |

mc-jev gives you two ways to use this: a browser UI with charts, or a plain terminal command.

Quick start

# install globally from npm
npm install -g mc-jev-cli

# start the web UI on port 7834
mc-jev serve -p 7834 --open

# run a request file and print a table in the terminal
mc-jev run request.json

# pipe JSON and get JSON back
cat request.json | mc-jev run - -f json

# list available Ollama models
mc-jev models

Set OLLAMA_HOST to override the default Ollama URL (http://localhost:11434).

CLI reference

mc-jev serve (default command)

Starts the Express + WebSocket server and opens the browser UI.

| Flag | Default | Description | |------|---------|-------------| | -p, --port <n> | 3737 | Port to listen on | | -u, --ollama <url> | $OLLAMA_HOST or http://localhost:11434 | Ollama base URL | | -m, --model <name> | nimble | Default model when the request omits model | | --open | off | Open the UI in your default browser |

mc-jev run [file]

Sends a request JSON file (or stdin with -) to Ollama and prints the answers.

| Flag | Default | Description | |------|---------|-------------| | -u, --ollama <url> | $OLLAMA_HOST | Ollama base URL | | -m, --model <name> | nimble | Model override | | -f, --format <type> | table | table (coloured terminal) or json |

mc-jev models

Lists models available on the Ollama server with their sizes.

Request format

{
  "model": "nimble",
  "state": {
    "ticket": "I was charged twice. Please refund the extra payment."
  },
  "questions": {
    "team": {
      "type": "choice",
      "instructions": "Which team should handle this ticket?",
      "criteria": {
        "billing": "Payments and refunds",
        "technical": "Bugs and integrations",
        "other": "None of the above"
      }
    },
    "refund": {
      "type": "noul",
      "instructions": "Does the customer explicitly ask for a refund?"
    },
    "urgency": {
      "type": "score",
      "instructions": "How urgent is this ticket?",
      "criteria": ["Routine", "Soon", "Urgent"]
    }
  }
}

Response format

{
  "model": "nimble",
  "answers": {
    "team": {
      "type": "choice",
      "choice": "billing",
      "probabilities": { "billing": 0.985, "technical": 0.012, "other": 0.003 },
      "confidence": 0.922
    },
    "refund": {
      "type": "noul",
      "noul": 0.997
    },
    "urgency": {
      "type": "score",
      "score": 0.815,
      "legend": { "0": "Routine", "1": "Soon", "2": "Urgent" },
      "probabilities": { "0": 0.378, "1": 0.429, "2": 0.193 },
      "confidence": 0.046
    }
  },
  "usage": { "input_tokens": 841, "output_tokens": 4 }
}

Built-in examples

The UI ships 10 ready-to-run examples covering:

1. Billing ticket triage
2. Product review sentiment
3. HR leave request
4. Support chat escalation
5. Bug report classification
6. Email spam detection
7. Job application screening
8. Content moderation
9. Medical symptom triage
10. Legal document risk

Select one from the — examples — dropdown in the toolbar, then press Ctrl+Enter to run it.

Keyboard shortcuts

| Shortcut | Action | |----------|--------| | Ctrl+Enter / Cmd+Enter | Run request | | Arrow keys on splitter | Resize panes |

Dependencies

| Package | Role | |---------|------| | express | HTTP server + static file serving | | ws | WebSocket server (real-time inference results) | | commander | CLI argument parsing | | monaco-editor | JSON editor in the browser (served locally) | | chart.js | Charts in the response panel (served locally) |

Node ≥ 18 required (uses native fetch).