feat(dev): Add image upload (vision) support for internal testing

[abeatrix]

Building block for #7235

This commit adds support for multi-media content in messages, particularly for image handling with byo dev models from google, including support for uploading images in the chat interface. Users can now upload images when they are using the configured model , and the uploaded images are then displayed as part of the chat messages as at-mention chip.

vision.mov

Key changes:

1. Added a content field to the Message interface with support for different message part types (text, media, tool)
2. Refactored Google Gemini message handling to properly process the new content structure
3. Updated chat client and sanitization logic to handle the new content field
4. Modified the UI components to conditionally show image upload only for BYOK and Vision models
5. Added Vision tag detection in the model creation function

Other changes:

• Adding a new MediaUploadButton component for handling image uploads.
• Adding a new ContextItemMedia type to represent media context items.
• Updating the constructGeminiChatMessages function to handle media content.
• Updating the renderContextItem function to handle media content.
• Updating the ContextItemMentionNode to render media context items.
• Updating the HumanMessageEditor to include the MediaUploadButton and handle media uploads.
• Adding @google/generative-ai as a dependency.
• Updating the Message type to include data and mimeType for media.
• Updating the chat-question telemetry event to include media context.
• Updating the googleChatClient to use the @google/generative-ai library.

Updated

Enterprise instances with Early Access feature flag enabled (e.g. sg02) could also enable vision models:

Test plan

Steps

For internal users: connected to sg02 instance

For BYOK

1. Add the following settings in your user settings.json file with your google ai studio key
2. In the options field, make sure to include "categories.vision" in the model that support "vision/image"
3. Build from this branch
4. You should not see the image upload icon when you are not on the cody.dev.models that has the "categories.vision" set

{
    "cody.dev.models": [
        {
            "provider": "google",
            "model": "gemini-2.0-flash-thinking-exp-01-21",
            "apiKey": "$GOOGLE_AI_STUDIO_KEY",
            "inputTokens": 1000000,
            "options": {
                "temperature": 1,
            }
        },
        {
            "provider": "google",
            "model": "gemini-2.0-flash",
            "apiKey": "$GOOGLE_AI_STUDIO_KEY",
            "inputTokens": 1000000,
            "options": {
                "temperature": 1,
                "categories": [
                    "vision" // <- This is required for the image icon to show up and enabled for the model
                ]
            },
        },
}

1. Take a screen shot of the home page of Cody
2. Copy and paste the screenshot into the editor
3. Ask "tell me what do you see"
4. Verify the answer

When sending to sg

[ykdojo]

Will be helpful to mention Ctrl+V for Windows/Linux users (maybe in a subsequent PR)

[ykdojo]

Will be helpful to mention Ctrl+V for Windows/Linux users (maybe in a subsequent PR)