Lark CLI v1.0.6: Multipart Uploads, VC Recording Shortcuts, and Concurrency Fixes
Lark CLI hit v1.0.6 today, just one day after v1.0.5. The pace has been fast — six versions in under two weeks since the initial open-source launch. Here is what is worth knowing about the latest drop, and why the multipart upload additions might be the most practically useful change so far.
Multipart uploads are finally here
Before v1.0.5 and v1.0.6, uploading large files through Lark CLI had a hard limit. Anything over 20MB would fail. For engineering teams sharing build artifacts, demo videos, or design files, this was a constant friction point.
v1.0.5 added multipart upload support for Drive files larger than 20MB. v1.0.6 extends this to document media uploads as well:
# Upload a large file to Drive
lark drive upload ./my-large-video.mp4
# Upload media to a Lark Doc
lark docs upload ./diagram.png --doc-token <token>
No manual chunking, no workarounds. The CLI handles the multipart logic automatically.
VC recording: a new shortcut
A useful addition in v1.0.6 is the +recording shortcut for video conference operations. Previously, converting a meeting_id to a minute_token — needed to access meeting recordings — required manual API calls. Now:
lark vc +recording --meeting-id <meeting_id>
One command returns the minute token, unlocking access to recordings, transcripts, and meeting summaries. For teams doing automated meeting-summary workflows, this is a clean building block.
Mail: local image paths just work now
When composing draft emails with embedded images, you can now reference local file paths directly — Lark CLI resolves and uploads them automatically across all draft entry points:
lark mail draft create \
--subject "Weekly Report" \
--body "./report-template.html" \
--to team@example.com
If report-template.html references local images, v1.0.6 auto-resolves those paths. No pre-uploading required.
Concurrency fix in RuntimeContext
A concurrency race condition in RuntimeContext was resolved. This was most likely to surface in scripts running multiple CLI commands in parallel, or in AI agent workflows invoking multiple Lark CLI tools simultaneously. If you run Lark CLI inside automated pipelines with parallel execution, this fix is worth the upgrade.
Task pagination from page token
The task module now supports starting pagination from an explicit page token:
lark task list --page-token <token>
For scheduled sync jobs that checkpoint their position, this makes it practical to resume pagination mid-stream without re-pulling the full list from the beginning.
Upgrade
# npm
npm install -g @larksuite/cli@latest
# Check version
lark --version
Full changelog: github.com/larksuite/cli