Fix: Allow SVG files as chat attachments (as code/text, not images)

[Copilot]

Fixes #228766

Problem

When dragging an SVG file into the chat input, it would show a yellow warning stating "does not support this file type" and prevent the file from being attached. This was problematic because SVG files are text-based XML documents that agents can edit just like any other text file.

Root Cause

The containsImageDragType() and extractImageFilesFromDragEvent() functions in chatDragAndDrop.ts were detecting SVG files as images (since their MIME type starts with image/), which caused them to go through the image handling path. However, SVG files should be treated as code/text files that agents can edit, not as binary images that models cannot understand.

Changes

Modified drag-and-drop image detection in src/vs/workbench/contrib/chat/browser/chatDragAndDrop.ts:

• Added isImageMimeType() helper function that excludes SVG files from image detection
• Updated containsImageDragType() to use the new helper, filtering out image/svg* MIME types
• Updated extractImageFilesFromDragEvent() to use the new helper

This allows SVG files to be handled as regular text/code files that can be attached and edited, while other image formats (PNG, JPEG, GIF, WebP) continue to be handled as images.

Impact

• ✅ SVG files can now be attached to chat without warnings
• ✅ SVG files are treated as text/code that agents can edit
• ✅ Models receive SVG as editable text content, not as image data
• ✅ Other image formats continue to work correctly as images
• ✅ Minimal, surgical changes to existing code

Fixes #270607

Original prompt

---

This section details on the original issue you should resolve

<issue_title>Can't attach an SVG to have the agent edit it</issue_title>
<issue_description>
Type: Bug

Drag SVG into chat input.
🐛 Yellow warning color and struck out as "… does not support this file type."

It is just text, and the agent could edit it just fine. Don't restrict it by treating it as unsupported image.

VS Code version: Code - Insiders 1.105.0-insider (Universal) (03c265b, 2025-10-08T14:07:41.558Z)
OS version: Darwin arm64 24.6.0
Modes:

System Info

Item	Value
CPUs	Apple M4 Pro (14 x 2400)
GPU Status	2d_canvas: enabled direct_rendering_display_compositor: disabled_off_ok gpu_compositing: enabled multiple_raster_threads: enabled_on opengl: enabled_on rasterization: enabled raw_draw: disabled_off_ok skia_graphite: enabled_on trees_in_viz: disabled_off video_decode: enabled video_encode: enabled webgl: enabled webgl2: enabled webgpu: enabled webnn: disabled_off
Load (avg)	9, 7, 6
Memory (System)	24.00GB (0.44GB free)
Process Argv	--log trace --enable-proposed-api digitarald.agent-memory --crash-reporter-id 9b206efe-c338-4f34-baff-2aab1c48a82e
Screen Reader	no
VM	0%

Comments on the Issue (you are @copilot in this section)

Fixes #270607

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.

[digitarald]

@copilot > The CHAT_ATTACHABLE_IMAGE_MIME_TYPES constant only included raster image formats (PNG, JPEG, GIF, WebP) but was missing SVG support.

But I don't want SVG to be handled as image, but as code. The models don't understand SVG as images anyways

[Copilot]

@copilot > The CHAT_ATTACHABLE_IMAGE_MIME_TYPES constant only included raster image formats (PNG, JPEG, GIF, WebP) but was missing SVG support.

But I don't want SVG to be handled as image, but as cod...

You're absolutely right. I've revised the approach to treat SVG files as code/text instead of images.

The fix now excludes SVG files from image detection in the drag-and-drop logic by adding a helper function that filters out image/svg* MIME types. This allows SVG files to be attached and edited as text content while other image formats (PNG, JPEG, etc.) continue to be handled as images.

Changes in commit 5c7a8b4.