Compress too large images for Google captions

This commit is contained in:
Cohee 2023-12-14 22:28:22 +02:00
parent cde9903fcb
commit 6406f76c7d
2 changed files with 12 additions and 3 deletions

View File

@ -22,15 +22,19 @@ export async function getMultimodalCaption(base64Img, prompt) {
throw new Error('MakerSuite API key is not set.');
}
// OpenRouter has a payload limit of ~2MB
// OpenRouter has a payload limit of ~2MB. Google is 4MB, but we love democracy.
const isGoogle = extension_settings.caption.multimodal_api === 'google';
const base64Bytes = base64Img.length * 0.75;
const compressionLimit = 2 * 1024 * 1024;
if (extension_settings.caption.multimodal_api === 'openrouter' && base64Bytes > compressionLimit) {
if (['google', 'openrouter'].includes(extension_settings.caption.multimodal_api) && base64Bytes > compressionLimit) {
const maxSide = 1024;
base64Img = await createThumbnail(base64Img, maxSide, maxSide, 'image/jpeg');
if (isGoogle) {
base64Img = base64Img.split(',')[1];
}
}
const isGoogle = extension_settings.caption.multimodal_api === 'google';
const apiResult = await fetch(`/api/${isGoogle ? 'google' : 'openai'}/caption-image`, {
method: 'POST',
headers: getRequestHeaders(),

View File

@ -1030,6 +1030,11 @@ export function loadFileToDocument(url, type) {
* @returns {Promise<string>} A promise that resolves to the thumbnail data URL.
*/
export function createThumbnail(dataUrl, maxWidth, maxHeight, type = 'image/jpeg') {
// Someone might pass in a base64 encoded string without the data URL prefix
if (!dataUrl.includes('data:')) {
dataUrl = `data:image/jpeg;base64,${dataUrl}`;
}
return new Promise((resolve, reject) => {
const img = new Image();
img.src = dataUrl;