Response format

How to read Madjik API payloads.

Latest-value responses

GET /v1/metrics and GET /v1/metrics/{id} return:

{
  "success": true,
  "data": [
    {
      "metric_id": "M10004",
      "metric_name": "USDT Peg Deviation",
      "value": -0.0403,
      "unit": "enterprise",
      "timestamp": "2026-09-10T07:59:13.469970Z",
      "category": "tether",
      "signal": "neutral",
      "confidence": 41.9,
      "computation_method": "classical",
      "computed_by": "Black Belt Labs",
      "fresh": true,
      "personalization": "public"
    }
  ],
  "count": 1,
  "timestamp": "2026-09-10T08:22:58.170248Z",
  "engine": "Black Belt Labs"
}

Field	Type	Description
`metric_id`	string	Metric id (e.g. `M10004`); ids are sequential and carry no meaning
`value`	number	Current raw value
`signal`	string	`bullish` / `bearish` / `neutral`
`confidence`	number	Engine confidence, 0–100
`computation_method`	string	`classical`, AI, or quantum
`fresh`	bool	Whether the value is inside its freshness SLA

Time series

GET /v1/metrics/{id}/timeseries points carry provenance and a normalized value; OHLC fields are present where the interval aggregates multiple points:

{
  "timestamp": "2026-09-09T12:00:00Z",
  "value": -0.0385,
  "value_normalized": 48.2,
  "provenance": "live",
  "open": -0.041, "high": -0.037, "low": -0.042, "close": -0.0385
}

Provenance is one of live, backfill_recomputed, backfill_approximate, backfill_model_retro, forecast — see Data provenance for exactly what each means.

The variant grid

GET /v1/metrics/{id}/variants returns the precomputed grid of change variants × time variants:

Change variants: absval (absolute value), abschg (absolute change), relchg (relative change)
Time variants: now, past1h, past4h, past8h, past24h, past7d, past30d — plus next24h / next7d / next30d where the metric's forecast beat its no-change baseline in out-of-sample backtesting. Where it did not, the forecast variant is deliberately absent rather than fabricated.

Multiple computation methods

GET /v1/metrics/{id}/methods returns every computation of the metric side by side — progr (classical), learn (AI), quantum (quantum simulator). Agreement across methods strengthens a signal; divergence is itself information.