Skip to content

fix(asr-worker): fix inference on audios as embedded documents#21

Merged
ClemDoum merged 1 commit intomainfrom
fix/embedded-document-handling
Apr 20, 2026
Merged

fix(asr-worker): fix inference on audios as embedded documents#21
ClemDoum merged 1 commit intomainfrom
fix/embedded-document-handling

Conversation

@ClemDoum
Copy link
Copy Markdown
Contributor

@ClemDoum ClemDoum commented Apr 20, 2026

Description

Inference on embedded documents failed because the Document.path of these documents is actually the path of the parent/root document.
In practice these files are located under: artifact_dir/project/id[0:2]/id[2:4]/raw

However inference can't run on file without extension because the file extension is used to handler file conversion into the inference file format.

Hence when running inference on embedded document we create a symlink to the original document with the proper file extension.

Changes

asr-worker

Fixed

  • propery handle inference on audio files stored as embedded files by creating a symlink to the worker workdir with the correct audio file extension

datashare-python

Added

  • added a FilesystemDocument object to reprensent a DS doc which can be stored on the original filesystem, stored as an artifact or copied into a workdir by a worker
  • added artifacts_root, workdir, docs_root to WorkerConfig

Changed

  • improved activity_contextual_id to be less verbose (only log wf info by default)

@ClemDoum ClemDoum self-assigned this Apr 20, 2026
@ClemDoum ClemDoum force-pushed the fix/embedded-document-handling branch from 735abbe to 3f41bc7 Compare April 20, 2026 15:04
@ClemDoum ClemDoum force-pushed the fix/embedded-document-handling branch from 3f41bc7 to d9d214f Compare April 20, 2026 15:16
@ClemDoum ClemDoum marked this pull request as ready for review April 20, 2026 15:17
@ClemDoum ClemDoum merged commit 8f1ebbd into main Apr 20, 2026
8 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant