You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This commit was created on GitHub.com and signed with GitHub’s verified signature.
v1.0.34
Vision Bridge v1.0.2
Removed
Filter valve skip_if_vision_capable. A text-only model must be marked vision-capable for Open WebUI to accept an image at all, so the valve disabled the filter in exactly the setup it exists for — the image reached the text-only model and returned an upstream 500. Enable the filter per model instead of globally.
Changed
Filter inlet no longer takes __model__ (unused after the valve removal).
Tool version bumped to 1.0.2 to stay in lockstep with the filter; no code change.
README rewritten for setup mistakes: five rules up front, a which-valve-goes-where table (both files have a vision_model_id, only one is read per mode), setup with exact UI sections plus a verification step, and a troubleshooting table mapping each error string to its cause and fix — including Model not found (the id must match Admin Panel → Settings → Models exactly, connection prefix included) and No matching image found in temporary chats.
README headings are plain text, no emoji.
Added
MCP App Bridge README: note that plain MCP servers belong in Open WebUI's native MCP support; the bridge is for MCP Apps that ship a ui:// resource.