Analytics

Aggregated trends across a user's calls, built from the per-call sentiment / tone fields already produced by analysis. Nothing here is a fresh LLM call — it's SQL aggregation over stored analysis results, so it's cheap to poll.

GET /api/analytics/vibe

Sentiment and tone distribution/trend for the authenticated user's calls.

Query parameters

Param	Type	Default	Notes
days	int	30	Look-back window in days.

Response — 200 OK

{
  "sentimentDistribution": { "positive": 12, "neutral": 5, "negative": 2, "mixed": 1 },
  "userSentimentDistribution": { "positive": 14, "neutral": 4, "negative": 2 },
  "contactSentimentDistribution": { "positive": 9, "neutral": 6, "negative": 5 },
  "sentimentTrend": [
    { "week": "2026-06-01", "counts": { "positive": 4, "neutral": 2, "negative": 1 } },
    { "week": "2026-06-08", "counts": { "positive": 5, "neutral": 1, "negative": 0 } }
  ],
  "toneProfile": {
    "formality": { "formal": 6, "semi-formal": 9, "informal": 3 },
    "energy": { "low": 2, "medium": 11, "high": 5 },
    "pace": { "slow": 1, "normal": 14, "fast": 3 }
  }
}

Fields

Field	Type	Notes
sentimentDistribution	map	Overall sentiment.overall value → call count, within the window.
userSentimentDistribution	map	Same, but for sentiment.userSentiment.overall (the phone owner's read).
contactSentimentDistribution	map	Same, for sentiment.contactSentiment.overall (the other party's read).
sentimentTrend	array	One point per ISO week in the window; counts is the same shape as sentimentDistribution.
toneProfile	object	Distribution of tone.formality / tone.energy / tone.pace across calls in the window.

Calls with no stored sentiment/tone (e.g. still processing, or pre-v2 analysis with only {overall, reason}) are excluded from the relevant distribution rather than counted as a bucket.

Errors

Status	Cause
401	Missing/invalid auth.

Storage

Populated from the call_sentiment_summary table, which is written alongside action_items when analysis side effects are applied — one row per completed call, denormalising just the fields this endpoint needs so the query doesn't have to deserialise every ANALYSIS_JSON artifact on every request.

Related

• Features · LLM analysis — where sentiment and tone come from.