Instinct endpoints

POST /v1/flag

Record a memory carrying caution weight. See flag.

curl -X POST https://mneva.dev/v1/flag \
  -H "x-mneva-key: $MNEVA_KEY" \
  -H "content-type: application/json" \
  -d '{
    "text": "<what is dangerous, and why>",
    "caution": 0.85
  }'

Field	Type	Required	Default
`text`	string	yes	—
`caution`	number 0..1	no	0.85

Response:

{ "id": 5, "text": "...", "caution": 0.85 }

Flags are persisted in the same memories table as remember, distinguished only by caution > 0. assess reads them.

POST /v1/assess

Gut-check a topic. See assess for full semantics and Instinct & Calibration for the concept.

curl -X POST https://mneva.dev/v1/assess \
  -H "x-mneva-key: $MNEVA_KEY" \
  -H "content-type: application/json" \
  -d '{"topic":"<what you are about to work on>"}'

Response (full shape, all fields):

{
  "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
    }]
  }
}

based_on is null when verdict is clear. track_record is null when no resolved predictions share a distinctive word with the topic. Neither block is ever fabricated.

Risk bands: >= 0.34 → danger, >= 0.12 → caution, else clear.