Merge pull request #3980 from SillyTavern/fix-lcpp-caption Fix llama.cpp captioning

b2ed69aac2b7c334c0b7d6d49657217188b0d412

Cohee <18619528+Cohee1207@users.noreply.github.com>

Signed

Parents: c6a64d85265b417683a8449766635222c6980524 aef005007f3b7810a90abb7d47106bb349344bfb

public/scripts/extensions/shared.js+1 -3 Viewed

     throwIfInvalidModel(useReverseProxy);
     const noPrefix = ['ollama', 'llamacpp'].includes(extension_settings.caption.multimodal_api);
     if (noPrefix && base64Img.startsWith('data:image/')) {
         base64Img = base64Img.split(',')[1];

                 return '/api/google/caption-image';
             case 'anthropic':
                 return '/api/anthropic/caption-image';
 -            case 'llamacpp':
 -                return '/api/backends/text-completions/llamacpp/caption-image';
             case 'ollama':
                 return '/api/backends/text-completions/ollama/caption-image';
             default:

src/endpoints/backends/text-completions.js+0 -45 Viewed

 const llamacpp = express.Router();
 -llamacpp.post('/caption-image', async function (request, response) {
 -    try {
 -        if (!request.body.server_url) {
 -            return response.sendStatus(400);
 -        }
+-
 -        console.debug('LlamaCpp caption request:', request.body);
 -        const baseUrl = trimV1(request.body.server_url);
+-
 -        const fetchResponse = await fetch(`${baseUrl}/completion`, {
 -            method: 'POST',
 -            headers: { 'Content-Type': 'application/json' },
 -            body: JSON.stringify({
 -                prompt: `USER:[img-1]${String(request.body.prompt).trim()}\nASSISTANT:`,
 -                image_data: [{ data: request.body.image, id: 1 }],
 -                temperature: 0.1,
 -                stream: false,
 -                stop: ['USER:', '</s>'],
 -            }),
 -        });
+-
 -        if (!fetchResponse.ok) {
 -            console.error('LlamaCpp caption error:', fetchResponse.status, fetchResponse.statusText);
 -            return response.status(500).send({ error: true });
 -        }
+-
 -        /** @type {any} */
 -        const data = await fetchResponse.json();
 -        console.debug('LlamaCpp caption response:', data);
+-
 -        const caption = data?.content || '';
+-
 -        if (!caption) {
 -            console.error('LlamaCpp caption is empty.');
 -            return response.status(500).send({ error: true });
 -        }
+-
 -        return response.send({ caption });
+-
 -    } catch (error) {
 -        console.error(error);
 -        return response.sendStatus(500);
 -    }
 -});
+-
 llamacpp.post('/props', async function (request, response) {
     try {
         if (!request.body.server_url) {

src/endpoints/openai.js+7 -2 Viewed

@@ -45,6 +45,10 @@ router.post('/caption-image', async (request, response) => {
45	45	key = readSecret(request.user.directories, SECRET_KEYS.KOBOLDCPP);
46	46	}
47	47
	48	+ if (request.body.api === 'llamacpp') {
	49	+ key = readSecret(request.user.directories, SECRET_KEYS.LLAMACPP);
	50	+ }
	51	+
48	52	if (request.body.api === 'vllm') {
49	53	key = readSecret(request.user.directories, SECRET_KEYS.VLLM);
50	54	}

@@ -69,7 +73,8 @@ router.post('/caption-image', async (request, response) => {
69	73	key = readSecret(request.user.directories, SECRET_KEYS.XAI);
70	74	}
71	75
72	76	if (!key &&const !request.body.reverse_proxynoKeyTypes &&= ['custom', 'ooba', 'koboldcpp', 'vllm'].includes(request.body.api) === false), {'llamacpp'];
	77	+ if (!key && !request.body.reverse_proxy && !noKeyTypes.includes(request.body.api)) {
73	78	console.warn('No key found for API', request.body.api);
74	79	return response.sendStatus(400);
75	80	}

@@ -156,7 +161,7 @@ router.post('/caption-image', async (request, response) => {
156	161	});
157	162	}
158	163
159	164	if (request.body.api === ['koboldcpp', \|\|'vllm', 'llamacpp'].includes(request.body.api === 'vllm')) {
160	165	apiUrl = `${trimV1(request.body.server_url)}/v1/chat/completions`;
161	166	}
162	167