Documentation

API reference and integration guides.

Authentication

All requests require your API key in the header:

X-API-Key: nobu_your_api_key_here

Keys are issued after payment. Dashboard services use the same key.

PowerSense API From $99/month (agent) or $15/hour (API)

Power quality monitoring. Two access modes: agent+dashboard (subscription, rate-limited to your tier) or direct API (usage-metered, unlimited rate).

{
  "device_id": "factory_main",
  "evidence": {
    "schema_id": "powersense_v1",
    "channels": ["vrms", "freq", "thd"],
    "series": [
      {"t_ms": 0, "v": [228.5, 49.92, 0.04]},
      {"t_ms": 1000, "v": [227.8, 49.88, 0.05]}
    ]
  }
}

Response

NOBU Desktop translates these fields into user-facing verdicts (Stable / Attention / Risk) and plain-language recommendations.

Predictive Maintenance API From $149/month per asset (agent) or $15/hour (API)

Equipment failure prediction. Agent+dashboard (subscription, rate-limited) or direct API (usage-metered, unlimited rate).

{
  "asset_id": "pump_north_1",
  "evidence": {
    "schema_id": "maintenance_v1",
    "channels": ["vibration", "temp", "current"],
    "series": [
      {"t_ms": 0, "v": [0.12, 45.2, 12.5]},
      {"t_ms": 1000, "v": [0.14, 45.3, 12.6]}
    ]
  }
}

Control API $15/hour

Optimal control for vehicles, robots, industrial systems.

{
  "observed": {
    "pos": 1.2,
    "vel": 0.5,
    "mode": 0.8
  }
}

Response

{
  "request_id": "req_xyz789",
  "meter": {"remaining_seconds": 3540},
  "action": {
    "u": [0.15, -0.02],
    "score": 0.87
  }
}

API Fuzzing $15/hour

Security testing that finds hidden states.

{
  "target_base_url": "https://api.example.com",
  "observed_responses": [
    {"method": "GET", "path": "/users", "status": 200},
    {"method": "POST", "path": "/users", "status": 401}
  ]
}

Anti-Distillation API $15/hour

Model distillation detection. Identifies systematic extraction of model capabilities from API query patterns. Detects both overt and evasive distillation attempts from API query patterns.

{
  "asset_id": "gpt4-production",
  "topic_novelty": 0.3,
  "structure_score": 0.7,
  "capability_depth": 0.8,
  "response_complexity": 0.6,
  "query_rate": 0.9,
  "pattern_entropy": 0.2,
  "coverage_delta": 0.85
}

Observation fields

Response

Stateful per asset_id: consecutive calls for the same asset accumulate evidence. Converges to 99%+ confidence within 5–8 observations for systematic distillers.

Errors

Cold Start Handling (202)

NOBU uses on-demand compute to minimize costs. If the compute engine is stopped, the first request returns 202 Accepted with a Retry-After: 5 header. The engine boots in 60–90 seconds. Your client should retry until it receives a 200.

// Check readiness or trigger warm-up:
GET /v1/compute/status
X-API-Key: your_key

// Response (warming):
// HTTP 202  { "ready": false, "state": "pending", "retry_after": 5 }

// Response (ready):
// HTTP 200  { "ready": true, "state": "running" }

You can call GET /v1/compute/status on startup to pre-warm the engine before submitting work. No billing is consumed for warm-up calls.

Billing

Agent + Dashboard (PowerSense, Maintenance subscriptions): Flat monthly rate. Inference calls are rate-limited to your tier's interval (5 minutes for Standard, 1 minute for Professional). No compute seconds are deducted. Access expires at end of billing period.

Direct API (PowerSense API, Maintenance API, Control, Fuzzing): Metered by compute time at $15/hour. No rate limit — call as fast as you want. Each response includes remaining balance. When exhausted, requests return 402. Subscription keys that exceed their rate limit receive 429 with a Retry-After header.

Top up your account →

tar xzf nobu-agent-linux.tar.gz
cd nobu-agent-linux
sudo bash install_linux.sh