`assess`

Gut-check a topic before working on it. Returns clear, caution, or danger with the reason — and, when the brain has resolved predictions in the adjacent area, the actual track record.

The most distinctive tool in Mneva. See Instinct & Calibration for the concept.

Signature

assess({ topic: string }) → {
  topic: string,
  verdict: 'clear' | 'caution' | 'danger',
  risk: number,
  why: string,
  based_on: null | {
    memory_id: number,
    caution: number,
    relevance: number,
    effective_relevance: number
  },
  track_record: null | {
    adjacent_predictions: number,
    correct: number,
    incorrect: number,
    partial: number,
    accuracy: number,
    mean_confidence: number,
    mean_surprise: number,
    most_surprising: Array<{ id, prediction, domain, outcome, surprise }>
  }
}

Param	Type	Required	Description
`topic`	string	yes	What you are about to work on.

Example

curl -X POST https://mneva.dev/v1/assess \
  -H "x-mneva-key: $MNEVA_KEY" \
  -H "content-type: application/json" \
  -d '{"topic":"deploy the new env variables to production"}'

Response from a brain with one flagged deploy memory + one resolved-incorrect deploy prediction:

{
  "topic": "deploy the new env variables to production",
  "verdict": "danger",
  "risk": 0.8,
  "why": "deploy failure: missing env var caused the last incident",
  "based_on": {
    "memory_id": 1,
    "caution": 0.8,
    "relevance": 0.6958,
    "effective_relevance": 1
  },
  "track_record": {
    "adjacent_predictions": 1,
    "correct": 0,
    "incorrect": 1,
    "partial": 0,
    "accuracy": 0,
    "mean_confidence": 0.9,
    "mean_surprise": 0.9,
    "most_surprising": [{
      "id": 1,
      "prediction": "deploy will succeed on first try",
      "domain": "deploy",
      "outcome": "incorrect",
      "surprise": 0.9
    }]
  }
}

Reading the verdict

verdict — clear (proceed), caution (slow down), danger (verify before acting).
risk — caution × effective_relevance, banded at >= 0.34 (danger) and >= 0.12 (caution).
why — the literal text of the flagged memory that drove the verdict. When verdict === 'clear', why is the constant string "Nothing wary known about this area."
based_on — non-null when verdict is non-clear. caution is the flagged memory's caution score; relevance is raw cosine; effective_relevance clamps to 1 when topic words overlap distinctively.
track_record — non-null when at least one resolved prediction is semantically adjacent to the topic (cosine of embeddings ≥ 0.35), or whose text shares a distinctive word with the topic as fallback. accuracy is correct / n. most_surprising lists predictions with surprise ≥ 0.5 (up to 3). The block is null when there is no signal, never invented.

assess

Signature

Example

Reading the verdict

See also

`assess`