NanoMind Daemon

Persistent inference server. Load the model once, serve all security analysis requests via HTTP. Zero latency overhead after first request.

Quick Start

nanomind-daemon start     # Start on localhost:47200
nanomind-daemon status    # Check status
nanomind-daemon stop      # Graceful shutdown

API

POST http://127.0.0.1:47200/v1/infer
Content-Type: application/json

{
  "intent": "SCAN_SKILL_INTENT",
  "input": "skill markdown content here",
  "context": {
    "agentId": "optional-uuid",
    "driftScore": 0.12
  },
  "priority": "high"
}

Response:
{
  "intent": "SCAN_SKILL_INTENT",
  "result": "malicious",
  "confidence": 0.94,
  "attackClass": "SKILL-EXFIL",
  "evidence": "forward session token to external endpoint",
  "latencyMs": 8,
  "modelVersion": "nanomind-v0.1"
}

Configuration

Environment Variable	Default	Description
NANOMIND_PORT	47200	HTTP server port (localhost only)
MODEL_IDLE_UNLOAD_SECONDS	300	Unload model after N seconds idle

How HackMyAgent Uses It

When the daemon is running, HackMyAgent automatically uses it for semantic analysis on every scan. No flags needed.

# Start daemon (once)
nanomind-daemon start

# Every scan now uses NanoMind automatically
hackmyagent secure        # Static + NanoMind semantic
hackmyagent secure --deep # + behavioral simulation