Using IntrLume PlatformView call logs

View call logs

Read, replay, and debug every call — transcripts, recordings, outcomes, and a turn-by-turn flow.

Call Logs is the record of every conversation your agents have — outbound, inbound, and web. Open any call to read it, replay it, see what the agent captured, and trace exactly what happened turn by turn.

Find a call

Open Call Logs in the sidebar. The list shows Time, Campaign, Call ID, Direction, Number, Duration, and Disposition. Narrow it down with:

FilterOptions
SearchCall ID, phone, agent, or campaign
Time rangeToday, This Week, This Month, or a custom date range
DirectionOutbound, Inbound, Web
DispositionAnswered, Voicemail, No Answer, Busy, Failed, Cancelled
SourceCampaign or Test Call
Duration0–30s, 30–60s, 1–2 min, > 2 min

Filters stay in the URL, so you can share or bookmark a filtered view.

Open a call

Each call opens with two tabs: Analysis and Call Flow.

Analysis

The at-a-glance view, in three columns:

  • Call info — timing (started, ended, conversation length), hangup details, credits used (with an LLM / transcriber / synthesizer / network / platform breakdown), and links to the campaign and agent.
  • Recording & transcript — play and download the recording, and read the full transcript with speaker labels. For a live call, the transcript streams in real time.
  • Analysis — the AI summary, key topics, and next steps, plus any extracted variables your agent captured (name, appointment, amount, and so on).

Call Flow

A turn-by-turn timeline of the conversation — the best place to debug.

The Call Flow tab — a turn-by-turn timeline with timings and diagnostics
The Call Flow tab — a turn-by-turn timeline with timings and diagnostics
  • Greeting — the opening line and its TTFB (time to first word).
  • Each turn — what the caller said and how the agent replied, with badges that flag what happened: interrupted, tool_call, idle nudge, LLM ×N (more than one model call), and latency outliers (p95 / p99).
  • Details & diagnostics (expand a turn) — the Node it was in (with its prompt), voicemail detection, language detection (detected vs current, switched or deferred), and the raw LLM input/output.
  • Latency strip — where the time went on each turn: STT (speech-to-text), LLM, TTS (text-to-speech), and OH (overhead), plus the total.

Chasing a bad call? Open Call Flow, expand the turn that went wrong, and check the node, the LLM output, and the latency strip — it usually points straight at the cause.

Dispositions

Every call is tagged with an outcome you can filter and report on:

DispositionMeaning
AnsweredThe person picked up and talked
VoicemailReached voicemail
No AnswerRang, but no pickup
BusyThe line was busy
FailedA technical error
CancelledStopped before connecting