INTERFACES

Commands, endpoints, and payloads for interacting with RUNE.

CLI Commands

python -m rune provides the following primary commands:

• run-ollama-instance: Provision or select an Ollama server.
• run-agentic-agent: Run HolmesGPT analysis.
• run-benchmark: Multi-phase workflow for provisioning + analysis.
• vastai-list-models: List configured Vast.ai model catalog.
• ollama-list-models: List models exposed by an existing Ollama server.

CLI Options Summary

Shared Agent Options

• --question, -q: The analysis question for HolmesGPT.
• --model, -m: The LLM model name to use.
• --kubeconfig: Path to the Kubernetes config file.
• --ollama-warmup, --no-ollama-warmup: Whether to pre-load the model before use.

Vast.ai Options (with --vastai)

• --vastai-template: The Vast.ai template hash.
• --vastai-min-dph, --vastai-max-dph: Dollars-per-hour range.
• --vastai-reliability: Minimum machine reliability percentage.

API Server (HTTP Mode)

Run the server with python -m rune.api.

Endpoints

• POST /v1/jobs/ollama: Submit an Ollama provisioning job.
• POST /v1/jobs/agent: Submit an agent analysis job.
• POST /v1/jobs/benchmark: Submit a full benchmark job.
• GET /v1/jobs/{job_id}: Poll for job status and results.
• GET /v1/models/vastai: List available Vast.ai models.
• GET /v1/models/ollama: List models from a target Ollama server.

Auth Headers

• Authorization: Bearer <token>
• X-API-Key: <key>
• X-Tenant-ID: <tenant>

REST API Reference

For a formal definition of all endpoints and request/reponse schemas, see the API Specification.

Wire Protocol (Driver Layer)

HolmesGPT is invoked via the DriverTransport layer.

Protocol JSON

Example ask action payload:

{
  "action": "ask",
  "params": {
    "question": "What's wrong?",
    "model": "llama3.1:8b",
    "kubeconfig": "..."
  }
}

The driver returns a structured JSON response with analysis results.