Skip to content
NewTalk to Basher from your browser — free

Platform · Analytics

Every call, measured where it ran.

The dashboard shows the operation as it happens — live calls, transcripts, outcomes — plus what each call actually cost and how fast it actually felt, from the same instrumentation the pipeline runs in production.

Illustration — a control room at night, an empty chair at a long console with one desk lamp lit on it, a gold line rising in a single peak on the screen wall above

The same screen, alive

Watch a day answer itself.

This is what a morning looks like on the dashboard: the day's calls tracing themselves hour by hour, every outcome finding its account, and the speed of every reply measured against the budget.

CALL-VOLUME · 24H

The day, as it happened

0006121824
Hour by hour across the demo workspace's busiest day — the morning wave, the midday lull, the afternoon return. The dashboard traces it as it happens.
OUTCOMES · 30D

Every call accounted for

  • Resolved
  • Escalated
  • Dropped
The circle closes only when every call has an outcome — the same three buckets the dashboard's own legend draws, nothing added. Escalation is a rule you wrote; a drop is named here rather than hidden.
TURN-LATENCY

How fast every reply landed

13 turns0100200300400500
300–400ms is the budget every release is gated on. The bars are the bench's 30 per-turn marks, taken by the pipeline's own tracker: 28 land inside the band, 1 comes in faster than it and 1 slower. The band was drawn around numbers already measured — not the other way round.

The day and the outcomes are drawn from the seeded demo workspace, in the dashboard's own three outcome buckets, the way the dashboard itself draws them; the latency band is bench-measured. Illustrative of the product's instrumentation, never customer results.

What the screen is built from

  • LIVE BOARD

    Calls as they happen

    Recent and live calls with real statuses, a per-call timeline, and the full transcript of every conversation — the same records the engine wrote while the call ran.

  • COST-BREAKDOWN

    The cost of a call, itemized

    Each call carries five cost fields — model, speech synthesis, transcription, telephony and total — accumulated as the call runs, then broken down per call and across date ranges.

  • TURN-LATENCY SLO

    Slowdowns you see before callers do

    You'll see calls slowing down before the first customer ever complains — because speed is measured on every single turn, never smoothed into an average. The moment the pace crosses the budget, the screen raises a flag; the turn-by-turn metrics live in the spec sheet below.

  • ANALYTICS API

    The same numbers, over REST

    Call analytics, agent performance, recent calls, per-call timelines, turn latency and cost breakdowns are served by the platform API. The dashboard consumes exactly what you can consume.

  • HONEST-EMPTY

    No number is invented

    Where a metric isn't measured yet, the screen says so. An empty state is information; a fabricated chart is not.

Spec sheet — as enforced in code

  • LLM · TTS · STT · telephony · total

    Cost fields per call

  • STT · LLM TTFT · TTS TTFB · E2E

    Turn metrics

  • p95 > budget

    SLO breach signal

  • last 200

    Turn values persisted per call

Turn metrics are the pipeline's own marks; long calls keep their most recent 200 turns of values, where degradation shows.

The platform

The rest of the platform

Hear the call you'll be measuring tomorrow.

Talk to the live agent, then picture these same minutes as transcripts, durations, and outcomes on your dashboard. Once you build your own agent, no picturing needed.

Talk to Basher liveBook a walkthrough