Skip to content

File maintenance jobs

Four utility jobs keep the audio file inventory healthy: relocating files to different storage, converting their format, splitting long files by silence, and verifying that every database record still has its file. All are regular jobs — filters, schedules, runs, and logs work as described in Jobs console, and each supports a test-only mode where applicable.

Relocate audio files

Administration › Storage › Relocate Audio Files — the Relocate audio files job physically moves recording files to another storage target and updates their database records. Typical uses: tiering old audio to cheaper storage (local disk → S3), or migrating off a retiring server.

Settings:

  • Destination storage target and Destination filename format (the parameterized template; post-processing parameters like %{tenant-name} are available).
  • If destination file exists — how to compare/handle an existing destination file.
  • Remove empty directories — clean up the source storage as files leave it.
  • Metadata file — write a JSON metadata file alongside each relocated audio file.
  • Decrypt files — decrypt during relocation (default keeps encrypted files encrypted; see Encryption).
  • Test only — log what would be moved without modifying data.

To make old paths resolve without moving anything, use file path rewrite rules instead — see Files location.

Convert audio files

Administration › Storage › Convert Audio Files — the Convert audio files job re-encodes existing recordings: audio file format (WAV/MP3), channel layout, sampling rate, and MP3 bitrate (8–256 kbps; 0 uses the default). The classic use is converting historical WAV archives to MP3 to reclaim storage. New recordings' format is set separately — see Audio format.

Split audio by silence

The Split audio by silence job cuts long recordings into separate conversations at silence boundaries, using voice-activity detection. It is used for ingestion paths that deliver continuous audio containing multiple conversations (for example, imported archives recorded as one file per line). VAD settings control the detection: speech threshold, analysis window size, minimum speech and silence durations, and speech padding. Push new calls to queue sends the resulting conversations into the processing pipeline (transcription and analysis); a test-only mode is available.

Check files

Check files verifies that the files referenced by database records actually exist and are readable on their storage. Choose the Check method by cost and thoroughness:

  • Check file size — existence and size only (fastest).
  • Read first 1024 bytes — confirms the file is readable.
  • Read full file — full read (slowest, most thorough).

Run it after storage migrations, path rewrites, or suspected storage incidents; failures are reported per record in the job's logs.