Assign users

[cameledev]

Purpose

Users

Proposal

Build upon: #741

• Use newly implemented VAD in metadata to associated user names to speech intervals. Use intersect these intervals with WhisperX output and compute score (overlap / speaker_duration). Assign name for largest score (if score > threshold).
• This method allows handling of multiple people for a same user (multilple people on same device).
• User the declared user name in transcription. If multiple people are detected for a user, also add the assigned user. (e.g: charlie (SPEAKER_2))
• Add tests for user_assign
• Some code restructuring was necessary

Example Output

Two speakers, single user/microphone (named Cameledev)

{
  "title": "Meeting \"qgd-fnzu-enl\" on 2026-04-22 at 10:15",
  "content": "\n*[Download your recording (external link)](http://localhost:3000/recording/dbd5ce3e-1905-48b0-8eb1-50ea0b20a369)*\n\n\n **Cameledev (SPEAKER_01)**:  Je parle avant que la trans... Je parle après que la transcription ait commencé. Cette fois-ci, je parle plus longtemps.\n\n **Cameledev (SPEAKER_00)**: Je parle avec une deuxième voix différente qui n'est pas la même que la première voix.\n\n **Cameledev (SPEAKER_01)**: Je reparle avec la première voix. Stop.",
  "email": "testmail@test.com",
  "sub": "042bdc02-9a87-436b-a3a3-9960c43ac425"
}

[cameledev]

What should we do with this ?

[cameledev]

Put back types. Can't recording_datetime be None ?

[cameledev]

use | None for consistency

[cameledev]

Didn't look into this

[sonarqubecloud]

Quality Gate passed

Issues
6 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud