Skip to content

Release 1.3.0: forward image inputs to claude - #5

Merged
rousan merged 1 commit into
mainfrom
develop
Jul 19, 2026
Merged

rousan merged 1 commit into
mainfrom
develop

Conversation

@rousan

@rousan rousan commented Jul 19, 2026

Copy link
Copy Markdown
Owner

Fixes image handling: image_url content parts were silently dropped, so vision requests reached claude with no picture. They're now extracted (both data: base64 and http(s) URLs) and forwarded as Anthropic image blocks via stream-json input. Text-only requests are unchanged.

Verified end-to-end (non-streaming, streaming, and text-only regression) against a real claude build.

Version bumped to 1.3.0 — merging triggers the tagged npm + GitHub release.

🤖 Generated with Claude Code

image_url content parts (data: base64 or http(s) URLs) were silently
dropped — only text was extracted — so vision requests reached claude
with no picture. Extract them into Anthropic image sources and, when any
are present, switch claude to stream-json input to deliver a single user
event with a text block plus image blocks. Text-only requests keep the
original text-input path unchanged.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
@rousan
rousan merged commit ace95c7 into main Jul 19, 2026
3 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