РусскийRUEnglishEN

Session Lifecycle

A session is the core unit of work in Whisperer: the time window during which the client listens to a meeting, transcribes speech, and receives LLM responses. Understanding the lifecycle helps you use start/pause correctly, interpret statuses in the dashboard, and troubleshoot issues.

When to Read This

Read this article if you want to understand:

  • what happens to your data at each phase of a meeting;
  • why a response is sometimes delayed and when the transcript is saved;
  • how the connection between the client and the server works.

Session Phases

1. Start

Pressing the "Start" button (or the keyboard shortcut) puts the client into active mode:

  1. The client opens a connection to the server and sends the device identifier and token.
  2. The server validates the subscription and minute quota. If the check passes, it confirms the session start and returns the session identifier.
  3. The client activates audio capture and begins recording.

2. Recording (Streaming)

Audio is transmitted in short chunks every ~0.8 seconds. Each chunk is labeled with the speaker: the user or the counterpart.

The server buffers the chunks and sends them for recognition. The transcript is returned to the client in real time and displayed in the LiveTranscriptStrip.

3. Request and Response

When the user presses "Ask" (or the keyboard shortcut), the client sends an ask frame with the current transcript context.

The server assembles the system prompt (role prompt → user context → RAG from knowledge base) and queries the LLM. The response streams in: text appears as it is generated and is displayed immediately in AnswerArea with Markdown support and pagination (if there are multiple responses).

4. Pause

Pressing "Pause" stops audio capture, but the WebSocket connection remains open. The session timer pauses. Resume by pressing "Resume".

5. End

Pressing "Stop" or closing the client:

  1. The client sends a session.end frame.
  2. The server flushes buffered transcripts and the list of completed responses to the database.
  3. The WebSocket closes.
  4. In the web dashboard, the session appears in the History section with the full transcript and analytics.

📸 [Screenshot: Whisperer panel in recording mode — LiveTranscriptStrip and control buttons visible]

Data Storage

By default, the transcript and LLM responses are saved to the database and are accessible in the dashboard. If no-logs mode is enabled, data is deleted after the session ends — but the minutes consumed are still counted.

Common Errors

Error Cause Fix
Session does not start Email not confirmed or no active subscription Check account status and quota
"Can't hear the other person" Screen recording permission not granted (macOS) or system audio capture unavailable (Windows) macOS Permissions / Windows
Transcript does not appear Poor internet connection or WebSocket loss Check your network; restart the session
LLM response takes too long AI provider is overloaded or a slow model is selected Switch the model in settings

Best Practices

  • Stop the session explicitly with the "Stop" button — this guarantees that data is written to history.
  • Pause during long breaks in the meeting — the timer stops and minutes are not wasted.
  • Use no-logs for confidential meetings where storing the transcript is undesirable.

Related Articles