jigor-cli — the jigor binary (CLI + HTTP gateway)

Ask System One decision models from the terminal or over HTTP. von and laya run locally as ONNX; jev runs remote on OpenRouter Decisions — one wire protocol across all of them: noul/choice/score questions in, typed answers out.

cargo install jigor-cli      # or: npm install --global @zatsepin/jigor  /  pip install jigor
jigor serve --host 127.0.0.1 --port 8000   # HTTP gateway: POST /v1/systemone, GET /healthz
jigor ask [--model <id>] [--provider <p>]  # the same wire over stdin, no HTTP layer
jigor models                               # provider x model pairs

The installed command is jigor. cargo install builds against the ONNX Runtime prebuilts, available for linux x64/arm64, macOS Apple Silicon and Windows x64; on Intel Macs use the prebuilt universal binary instead (uvx jigor, pip install jigor or npm install --global @zatsepin/jigor).

Model aliases: von, laya, jev, jev-latest. The local ONNX models download to ~/.cache/huggingface on first use; set OPENROUTER_API_KEY for the remote jev backend.

Example

jigor ask --model von <<'JSON'
{
  "state": { "error": "Disk volume /var/log at 98% capacity." },
  "questions": {
    "requires_intervention": {
      "type": "noul",
      "instructions": "Does this disk space condition require operational intervention?"
    }
  }
}
JSON
# {"model":"von-1.0.0","backend":"local","answers":{"requires_intervention":{"type":"noul","noul":0.2739}}}

The jigor Rust library

This binary is a thin shell over the jigor crate — local ONNX backends in von/laya, remote jev via OpenRouter Decisions, and one answers() call for all three. Use the library directly when you want decisions inside a Rust program: crate docs · full project README.