Hello Chaiyapol,
Thanks for the questions. I was able to find information on both the archival process and recording exports that should help with planning.
Large-Scale Archival Activities
Genesys load testing indicates that the job creation phase for a bulk archive operation can take up to 1 hour for approximately 500,000 conversations within a single organization. Processing time scales with the number of conversations being archived, and the actual archive execution may take additional time beyond the initial job creation.
What I wasn't able to find in the public documentation is specific guidance around operational impacts during archive execution, such as:
Effects on recording playback or other real-time operations
API response time degradation
Recommended maintenance windows
Throttling strategies or ideal batch sizes
Because those details aren't currently documented, I'd recommend validating with smaller test batches first and monitoring performance in your environment before running a large-scale archive. If you're planning a significant archival effort, Genesys Support may also be able to provide additional guidance based on your organization's specific configuration and data volumes.
Recording Export Format and Metadata
For both the Recording Bulk Jobs API (EXPORT action) and automatic recording export policies, the exported output is structured the same way.
Recording formats:
Audio recordings: .opus by default (recording policies may also allow MP3, WAV, or WEBM)
Screen recordings: .zip files containing WEBM content
Digital recordings: HTML files
Metadata:
Each exported recording includes a corresponding _metadata.json file containing information such as:
Recording, conversation, and organization IDs
Start time, end time, and duration
Media type and subtype
ANI/DNIS information
User IDs, queue IDs, and wrap-up codes
Direction, language, file size, and file path
Files are organized in S3 using the following folder structure:
organization_id/year/month/day/hour/conversation_id/
Transcript Availability
One item worth calling out: transcripts are not automatically included in recording exports. If transcription data is needed, it must be retrieved separately through the Speech and Text Analytics APIs. Transcript retention follows the same QM retention policy as the associated recordings.
Summary
Whether you use the Bulk Jobs API or automatic QM export policies, you can expect the same export structure, metadata format, and S3 organization. The main difference is simply how and when the export is triggered.
For the archival side, the documented scaling guidance provides a good baseline, but I haven't found published recommendations around performance impact, throttling, or maintenance windows. If anyone in the community has practical experience running archives at this scale, it would be great to hear about your results as well.
Hope this helps!
------------------------------
Cameron
Online Community Manager/Moderator
------------------------------