rixty/SillyTavern
1
0
Fork 0
mirror of https://github.com/SillyTavern/SillyTavern.git synced 2025-06-05 21:59:27 +02:00
Code Issues Packages Projects Releases Wiki Activity

Add multimodal captioning for ooba

Browse Source
This commit is contained in:
Cohee
2023-12-24 01:43:29 +02:00
parent 88993bd3e8
commit a8fb306c12
3 changed files with 34 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
public/scripts/extensions/shared.js
View File

@ -21,13 +21,15 @@ export async function getMultimodalCaption(base64Img, prompt) {
}
// OpenRouter has a payload limit of ~2MB. Google is 4MB, but we love democracy.
// Ooba requires all images to be JPEGs.
const isGoogle = extension_settings.caption.multimodal_api === 'google';
const isOllama = extension_settings.caption.multimodal_api === 'ollama';
const isLlamaCpp = extension_settings.caption.multimodal_api === 'llamacpp';
const isCustom = extension_settings.caption.multimodal_api === 'custom';
const isOoba = extension_settings.caption.multimodal_api === 'ooba';
const base64Bytes = base64Img.length * 0.75;
const compressionLimit = 2 * 1024 * 1024;
if (['google', 'openrouter'].includes(extension_settings.caption.multimodal_api) && base64Bytes > compressionLimit) {
if ((['google', 'openrouter'].includes(extension_settings.caption.multimodal_api) && base64Bytes > compressionLimit) || isOoba) {
const maxSide = 1024;
base64Img = await createThumbnail(base64Img, maxSide, maxSide, 'image/jpeg');
@ -69,6 +71,10 @@ export async function getMultimodalCaption(base64Img, prompt) {
requestBody.server_url = textgenerationwebui_settings.server_urls[textgen_types.LLAMACPP];
}
if (isOoba) {
requestBody.server_url = textgenerationwebui_settings.server_urls[textgen_types.OOBA];
}
if (isCustom) {
requestBody.server_url = oai_settings.custom_url;
requestBody.model = oai_settings.custom_model || 'gpt-4-vision-preview';
@ -129,6 +135,10 @@ function throwIfInvalidModel() {
throw new Error('LlamaCPP server URL is not set.');
}
if (extension_settings.caption.multimodal_api === 'ooba' && !textgenerationwebui_settings.server_urls[textgen_types.OOBA]) {
throw new Error('Text Generation WebUI server URL is not set.');
}
if (extension_settings.caption.multimodal_api === 'custom' && !oai_settings.custom_url) {
throw new Error('Custom API URL is not set.');
}