mirror of
https://github.com/SillyTavern/SillyTavern.git
synced 2025-06-05 21:59:27 +02:00
Fix sending PNG/WEBP to Google captioning
This commit is contained in:
@ -1756,6 +1756,7 @@ async function generateMultimodalPrompt(generationType, quietPrompt) {
|
||||
}
|
||||
}
|
||||
|
||||
try {
|
||||
const response = await fetch(avatarUrl);
|
||||
|
||||
if (!response.ok) {
|
||||
@ -1768,10 +1769,15 @@ async function generateMultimodalPrompt(generationType, quietPrompt) {
|
||||
const caption = await getMultimodalCaption(avatarBase64, quietPrompt);
|
||||
|
||||
if (!caption) {
|
||||
throw new Error('Multimodal captioning failed.');
|
||||
throw new Error('No caption returned from the API.');
|
||||
}
|
||||
|
||||
return caption;
|
||||
} catch (error) {
|
||||
console.error(error);
|
||||
toastr.error('Multimodal captioning failed. Please try again.', 'Image Generation');
|
||||
throw new Error('Multimodal captioning failed.');
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
|
@ -8,14 +8,17 @@ const router = express.Router();
|
||||
|
||||
router.post('/caption-image', jsonParser, async (request, response) => {
|
||||
try {
|
||||
const mimeType = request.body.image.split(';')[0].split(':')[1];
|
||||
const base64Data = request.body.image.split(',')[1];
|
||||
const url = `https://generativelanguage.googleapis.com/v1beta/models/gemini-pro-vision:generateContent?key=${readSecret(SECRET_KEYS.MAKERSUITE)}`;
|
||||
const body = {
|
||||
contents: [{
|
||||
parts: [
|
||||
{ text: request.body.prompt },
|
||||
{ inlineData: {
|
||||
mimeType: 'image/png', //jpg images seem to work fine even with this mimetype set?
|
||||
data: request.body.image,
|
||||
{
|
||||
inlineData: {
|
||||
mimeType: 'image/png', // It needs to specify a MIME type in data if it's not a PNG
|
||||
data: mimeType === 'image/png' ? base64Data : request.body.image,
|
||||
},
|
||||
}],
|
||||
}],
|
||||
@ -23,6 +26,8 @@ router.post('/caption-image', jsonParser, async (request, response) => {
|
||||
generationConfig: { maxOutputTokens: 1000 },
|
||||
};
|
||||
|
||||
console.log('Multimodal captioning request', body);
|
||||
|
||||
const result = await fetch(url, {
|
||||
body: JSON.stringify(body),
|
||||
method: 'POST',
|
||||
@ -32,10 +37,9 @@ router.post('/caption-image', jsonParser, async (request, response) => {
|
||||
timeout: 0,
|
||||
});
|
||||
|
||||
console.log('Multimodal captioning request', body);
|
||||
|
||||
if (!result.ok) {
|
||||
console.log(`MakerSuite API returned error: ${result.status} ${result.statusText} ${await result.text()}`);
|
||||
const error = await result.json();
|
||||
console.log(`MakerSuite API returned error: ${result.status} ${result.statusText}`, error);
|
||||
return response.status(result.status).send({ error: true });
|
||||
}
|
||||
|
||||
|
Reference in New Issue
Block a user