Groq: Add new models and multimodal captions

6706cce10de9cc02e2337d512120ec3a0e8d431c

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

Parents: 85b0d135f1d709f92405566ba2e801dcf98ff28b
6 files changed, +47 -12Ignore whitespace
public/index.html+25 -10
@@ -2954,16 +2954,31 @@
29542954 </div>
29552955 <h4 data-i18n="Groq Model">Groq Model</h4>
29562956 <select id="model_groq_select">
2957- <option value="llama-3.1-405b-reasoning">llama-3.1-405b-reasoning</option>
2957+ <optgroup label="Llama 3.2">
29582958 <option value="llama-3.12-70b1b-versatilepreview">llama-3.12-70b1b-versatilepreview</option>
29592959 <option value="llama-3.12-8b3b-instantpreview">llama-3.12-8b3b-instantpreview</option>
29602960 <option value="llama3-groq-70bllama-81923.2-tool11b-usevision-preview">llama3-groq-70bllama-81923.2-tool11b-usevision-preview</option>
29612961 <option value="llama3-groq-8bllama-81923.2-tool90b-usevision-preview">llama3-groq-8bllama-81923.2-tool90b-usevision-preview</option>
2962- <option value="llama3-8b-8192">llama3-8b-8192</option>
2962+ </optgroup>
2963- <option value="llama3-70b-8192">llama3-70b-8192</option>
2963+ <optgroup label="Llama 3.1">
29642964 <option value="mixtralllama-8x7b3.1-327688b-instant">mixtralllama-8x7b3.1-327688b-instant</option>
29652965 <option value="gemmallama-7b3.1-it70b-versatile">gemmallama-7b3.1-it70b-versatile</option>
29662966 <option value="gemma2llama-9b3.1-it405b-reasoning">gemma2llama-9b3.1-it405b-reasoning</option>
2967+ </optgroup>
2968+ <optgroup label="Llama 3">
2969+ <option value="llama3-groq-8b-8192-tool-use-preview">llama3-groq-8b-8192-tool-use-preview</option>
2970+ <option value="llama3-groq-70b-8192-tool-use-preview">llama3-groq-70b-8192-tool-use-preview</option>
2971+ <option value="llama3-8b-8192">llama3-8b-8192</option>
2972+ <option value="llama3-70b-8192">llama3-70b-8192</option>
2973+ </optgroup>
2974+ <optgroup label="Gemma">
2975+ <option value="gemma-7b-it">gemma-7b-it</option>
2976+ <option value="gemma2-9b-it">gemma2-9b-it</option>
2977+ </optgroup>
2978+ <optgroup label="Other">
2979+ <option value="mixtral-8x7b-32768">mixtral-8x7b-32768</option>
2980+ <option value="llava-v1.5-7b-4096-preview">llava-v1.5-7b-4096-preview</option>
2981+ </optgroup>
29672982 </select>
29682983 </div>
29692984 <div id="perplexity_form" data-source="perplexity">
public/scripts/extensions/caption/index.js+1 -0
@@ -403,6 +403,7 @@ jQuery(async function () {
403403 (extension_settings.caption.source === 'multimodal' && extension_settings.caption.multimodal_api === 'openai' && (secret_state[SECRET_KEYS.OPENAI] || extension_settings.caption.allow_reverse_proxy)) ||
404404 (extension_settings.caption.source === 'multimodal' && extension_settings.caption.multimodal_api === 'openrouter' && secret_state[SECRET_KEYS.OPENROUTER]) ||
405405 (extension_settings.caption.source === 'multimodal' && extension_settings.caption.multimodal_api === 'zerooneai' && secret_state[SECRET_KEYS.ZEROONEAI]) ||
406+ (extension_settings.caption.source === 'multimodal' && extension_settings.caption.multimodal_api === 'groq' && secret_state[SECRET_KEYS.GROQ]) ||
406407 (extension_settings.caption.source === 'multimodal' && extension_settings.caption.multimodal_api === 'mistral' && (secret_state[SECRET_KEYS.MISTRALAI] || extension_settings.caption.allow_reverse_proxy)) ||
407408 (extension_settings.caption.source === 'multimodal' && extension_settings.caption.multimodal_api === 'google' && (secret_state[SECRET_KEYS.MAKERSUITE] || extension_settings.caption.allow_reverse_proxy)) ||
408409 (extension_settings.caption.source === 'multimodal' && extension_settings.caption.multimodal_api === 'anthropic' && (secret_state[SECRET_KEYS.CLAUDE] || extension_settings.caption.allow_reverse_proxy)) ||
public/scripts/extensions/caption/settings.html+4 -0
@@ -21,6 +21,7 @@
2121 <option value="anthropic">Anthropic</option>
2222 <option value="custom" data-i18n="Custom (OpenAI-compatible)">Custom (OpenAI-compatible)</option>
2323 <option value="google">Google AI Studio</option>
24+ <option value="groq">Groq</option>
2425 <option value="koboldcpp">KoboldCpp</option>
2526 <option value="llamacpp">llama.cpp</option>
2627 <option value="mistral">MistralAI</option>
@@ -60,6 +61,9 @@
6061 <option data-type="google" value="gemini-1.5-pro-exp-0801">gemini-1.5-pro-exp-0801</option>
6162 <option data-type="google" value="gemini-1.5-pro-exp-0827">gemini-1.5-pro-exp-0827</option>
6263 <option data-type="google" value="gemini-pro-vision">gemini-pro-vision</option>
64+ <option data-type="groq" value="llama-3.2-11b-vision-preview">llama-3.2-11b-vision-preview</option>
65+ <option data-type="groq" value="llama-3.2-90b-vision-preview">llama-3.2-90b-vision-preview</option>
66+ <option data-type="groq" value="llava-v1.5-7b-4096-preview">llava-v1.5-7b-4096-preview</option>
6367 <option data-type="openrouter" value="openai/gpt-4-vision-preview">openai/gpt-4-vision-preview</option>
6468 <option data-type="openrouter" value="openai/gpt-4o">openai/gpt-4o</option>
6569 <option data-type="openrouter" value="openai/gpt-4o-2024-05-13">openai/gpt-4o-2024-05-13</option>
public/scripts/extensions/shared.js+5 -1
@@ -36,7 +36,7 @@ export async function getMultimodalCaption(base64Img, prompt) {
3636 const isVllm = extension_settings.caption.multimodal_api === 'vllm';
3737 const base64Bytes = base64Img.length * 0.75;
3838 const compressionLimit = 2 * 1024 * 1024;
3939 if ((['google', 'openrouter', 'mistral', 'groq'].includes(extension_settings.caption.multimodal_api) && base64Bytes > compressionLimit) || isOoba || isKoboldCpp) {
4040 const maxSide = 1024;
4141 base64Img = await createThumbnail(base64Img, maxSide, maxSide, 'image/jpeg');
4242 }
@@ -135,6 +135,10 @@ function throwIfInvalidModel(useReverseProxy) {
135135 throw new Error('01.AI API key is not set.');
136136 }
137137
138+ if (extension_settings.caption.multimodal_api === 'groq' && !secret_state[SECRET_KEYS.GROQ]) {
139+ throw new Error('Groq API key is not set.');
140+ }
141+
138142 if (extension_settings.caption.multimodal_api === 'google' && !secret_state[SECRET_KEYS.MAKERSUITE] && !useReverseProxy) {
139143 throw new Error('Google AI Studio API key is not set.');
140144 }
public/scripts/openai.js+4 -1
@@ -4196,7 +4196,10 @@ async function onModelChange() {
41964196 if (oai_settings.max_context_unlocked) {
41974197 $('#openai_max_context').attr('max', unlocked_max);
41984198 }
41994199 else if (oai_settings.groq_model.includes('llama-3.12') && oai_settings.groq_model.includes('-preview')) {
4200+ $('#openai_max_context').attr('max', max_8k);
4201+ }
4202+ else if (oai_settings.groq_model.includes('llama-3.2') || oai_settings.groq_model.includes('llama-3.1')) {
42004203 $('#openai_max_context').attr('max', max_128k);
42014204 }
42024205 else if (oai_settings.groq_model.includes('llama3-groq')) {
src/endpoints/openai.js+8 -0
@@ -55,6 +55,10 @@ router.post('/caption-image', jsonParser, async (request, response) => {
5555 key = readSecret(request.user.directories, SECRET_KEYS.MISTRALAI);
5656 }
5757
58+ if (request.body.api === 'groq') {
59+ key = readSecret(request.user.directories, SECRET_KEYS.GROQ);
60+ }
61+
5862 if (!key && !request.body.reverse_proxy && ['custom', 'ooba', 'koboldcpp', 'vllm'].includes(request.body.api) === false) {
5963 console.log('No key found for API', request.body.api);
6064 return response.sendStatus(400);
@@ -111,6 +115,10 @@ router.post('/caption-image', jsonParser, async (request, response) => {
111115 apiUrl = 'https://api.01.ai/v1/chat/completions';
112116 }
113117
118+ if (request.body.api === 'groq') {
119+ apiUrl = 'https://api.groq.com/openai/v1/chat/completions';
120+ }
121+
114122 if (request.body.api === 'mistral') {
115123 apiUrl = 'https://api.mistral.ai/v1/chat/completions';
116124 }