SillyTavern/public/scripts/extensions/caption/index.js

317 lines
11 KiB
JavaScript
Raw Normal View History

import { getBase64Async, saveBase64AsFile } from "../../utils.js";
import { getContext, getApiUrl, doExtrasFetch, extension_settings, modules } from "../../extensions.js";
2023-11-07 00:58:34 +01:00
import { callPopup, getRequestHeaders, saveSettingsDebounced, substituteParams } from "../../../script.js";
import { getMessageTimeStamp } from "../../RossAscends-mods.js";
import { SECRET_KEYS, secret_state } from "../../secrets.js";
2023-07-20 19:32:15 +02:00
export { MODULE_NAME };
const MODULE_NAME = 'caption';
const UPDATE_INTERVAL = 1000;
2023-11-07 00:58:34 +01:00
const PROMPT_DEFAULT = 'Whats in this image?';
const TEMPLATE_DEFAULT = '[{{user}} sends {{char}} a picture that contains: {{caption}}]';
2023-07-20 19:32:15 +02:00
async function moduleWorker() {
const hasConnection = getContext().onlineStatus !== 'no_connection';
$('#send_picture').toggle(hasConnection);
2023-07-20 19:32:15 +02:00
}
2023-11-07 00:58:34 +01:00
function migrateSettings() {
if (extension_settings.caption.local !== undefined) {
extension_settings.caption.source = extension_settings.caption.local ? 'local' : 'extras';
}
delete extension_settings.caption.local;
2023-11-07 00:58:34 +01:00
if (!extension_settings.caption.source) {
extension_settings.caption.source = 'extras';
}
if (!extension_settings.caption.prompt) {
extension_settings.caption.prompt = PROMPT_DEFAULT;
}
if (!extension_settings.caption.template) {
extension_settings.caption.template = TEMPLATE_DEFAULT;
}
}
2023-07-20 19:32:15 +02:00
async function setImageIcon() {
try {
const sendButton = $('#send_picture .extensionsMenuExtensionButton');
sendButton.addClass('fa-image');
sendButton.removeClass('fa-hourglass-half');
}
catch (error) {
console.log(error);
}
}
async function setSpinnerIcon() {
try {
const sendButton = $('#send_picture .extensionsMenuExtensionButton');
sendButton.removeClass('fa-image');
sendButton.addClass('fa-hourglass-half');
}
catch (error) {
console.log(error);
}
}
async function sendCaptionedMessage(caption, image) {
const context = getContext();
2023-11-07 00:58:34 +01:00
let template = extension_settings.caption.template || TEMPLATE_DEFAULT;
if (!/{{caption}}/i.test(template)) {
console.warn('Poka-yoke: Caption template does not contain {{caption}}. Appending it.')
template += ' {{caption}}';
}
let messageText = substituteParams(template).replace(/{{caption}}/i, caption);
2023-07-20 19:32:15 +02:00
if (extension_settings.caption.refine_mode) {
messageText = await callPopup(
'<h3>Review and edit the generated message:</h3>Press "Cancel" to abort the caption sending.',
'input',
messageText,
{ rows: 5, okButton: 'Send' });
if (!messageText) {
throw new Error('User aborted the caption sending.');
}
}
const message = {
name: context.name1,
is_user: true,
send_date: getMessageTimeStamp(),
2023-07-20 19:32:15 +02:00
mes: messageText,
extra: {
image: image,
title: messageText,
},
};
context.chat.push(message);
context.addOneMessage(message);
await context.generate('caption');
}
/**
*
* @param {string} base64Img Base64 encoded image without the data:image/...;base64, prefix
* @param {string} fileData Base64 encoded image with the data:image/...;base64, prefix
* @returns
*/
async function doCaptionRequest(base64Img, fileData) {
switch (extension_settings.caption.source) {
case 'local':
return await captionLocal(base64Img);
case 'extras':
return await captionExtras(base64Img);
case 'horde':
return await captionHorde(base64Img);
case 'openai':
return await captionOpenAI(fileData);
default:
throw new Error('Unknown caption source.');
}
}
2023-07-20 19:32:15 +02:00
async function captionExtras(base64Img) {
if (!modules.includes('caption')) {
throw new Error('No captioning module is available.');
}
2023-07-20 19:32:15 +02:00
const url = new URL(getApiUrl());
url.pathname = '/api/caption';
2023-07-20 19:32:15 +02:00
const apiResult = await doExtrasFetch(url, {
method: 'POST',
headers: {
'Content-Type': 'application/json',
'Bypass-Tunnel-Reminder': 'bypass',
},
body: JSON.stringify({ image: base64Img })
});
if (!apiResult.ok) {
throw new Error('Failed to caption image via Extras.');
}
const data = await apiResult.json();
return data;
}
async function captionLocal(base64Img) {
const apiResult = await fetch('/api/extra/caption', {
method: 'POST',
headers: getRequestHeaders(),
body: JSON.stringify({ image: base64Img })
});
if (!apiResult.ok) {
throw new Error('Failed to caption image via local pipeline.');
}
const data = await apiResult.json();
return data;
}
async function captionHorde(base64Img) {
const apiResult = await fetch('/api/horde/caption-image', {
method: 'POST',
headers: getRequestHeaders(),
body: JSON.stringify({ image: base64Img })
});
if (!apiResult.ok) {
throw new Error('Failed to caption image via Horde.');
}
const data = await apiResult.json();
return data;
}
async function captionOpenAI(base64Img) {
2023-11-07 00:58:34 +01:00
const prompt = extension_settings.caption.prompt || PROMPT_DEFAULT;
const apiResult = await fetch('/api/openai/caption-image', {
method: 'POST',
headers: getRequestHeaders(),
2023-11-07 00:58:34 +01:00
body: JSON.stringify({ image: base64Img, prompt: prompt }),
});
if (!apiResult.ok) {
throw new Error('Failed to caption image via OpenAI.');
}
const data = await apiResult.json();
return data;
}
async function onSelectImage(e) {
setSpinnerIcon();
const file = e.target.files[0];
if (!file || !(file instanceof File)) {
return;
}
try {
const context = getContext();
const fileData = await getBase64Async(file);
const base64Format = fileData.split(',')[0].split(';')[0].split('/')[1];
const base64Data = fileData.split(',')[1];
const data = await doCaptionRequest(base64Data, fileData);
const caption = data.caption;
const imageToSave = data.thumbnail ? data.thumbnail : base64Data;
const format = data.thumbnail ? 'jpeg' : base64Format;
const imagePath = await saveBase64AsFile(imageToSave, context.name2, '', format);
await sendCaptionedMessage(caption, imagePath);
2023-07-20 19:32:15 +02:00
}
catch (error) {
toastr.error('Failed to caption image.');
2023-07-20 19:32:15 +02:00
console.log(error);
}
finally {
e.target.form.reset();
setImageIcon();
}
}
function onRefineModeInput() {
extension_settings.caption.refine_mode = $('#caption_refine_mode').prop('checked');
saveSettingsDebounced();
}
jQuery(function () {
function addSendPictureButton() {
const sendButton = $(`
<div id="send_picture" class="list-group-item flex-container flexGap5">
<div class="fa-solid fa-image extensionsMenuExtensionButton"></div>
Send a Picture
2023-07-20 19:32:15 +02:00
</div>`);
$('#extensionsMenu').prepend(sendButton);
$(sendButton).hide();
$(sendButton).on('click', () => {
const hasCaptionModule =
(modules.includes('caption') && extension_settings.caption.source === 'extras') ||
(extension_settings.caption.source === 'openai' && secret_state[SECRET_KEYS.OPENAI]) ||
extension_settings.caption.source === 'local' ||
extension_settings.caption.source === 'horde';
if (!hasCaptionModule) {
2023-11-07 00:58:34 +01:00
toastr.error('No captioning module is available. Choose other captioning source in the extension settings.');
return;
}
$('#img_file').trigger('click');
});
2023-07-20 19:32:15 +02:00
}
function addPictureSendForm() {
const inputHtml = `<input id="img_file" type="file" accept="image/*">`;
const imgForm = document.createElement('form');
imgForm.id = 'img_form';
$(imgForm).append(inputHtml);
$(imgForm).hide();
$('#form_sheld').append(imgForm);
$('#img_file').on('change', onSelectImage);
}
function addSettings() {
const html = `
<div class="caption_settings">
2023-07-20 19:32:15 +02:00
<div class="inline-drawer">
<div class="inline-drawer-toggle inline-drawer-header">
<b>Image Captioning</b>
<div class="inline-drawer-icon fa-solid fa-circle-chevron-down down"></div>
</div>
<div class="inline-drawer-content">
<label for="caption_source">Source:</label>
2023-11-07 00:58:34 +01:00
<select id="caption_source" class="text_pole">
<option value="local">Local</option>
<option value="extras">Extras</option>
<option value="horde">Horde</option>
<option value="openai">OpenAI</option>
</select>
2023-11-07 00:58:34 +01:00
<label for="caption_prompt">Caption Prompt (OpenAI):</label>
<textarea id="caption_prompt" class="text_pole" rows="1" placeholder="&lt; Use default &gt;">${PROMPT_DEFAULT}</textarea>
<label for="caption_template">Message Template: <small>(use <tt>{{caption}}</tt> macro)</small></label>
<textarea id="caption_template" class="text_pole" rows="2" placeholder="&lt; Use default &gt;">${TEMPLATE_DEFAULT}</textarea>
<label class="checkbox_label margin-bot-10px" for="caption_refine_mode">
2023-07-20 19:32:15 +02:00
<input id="caption_refine_mode" type="checkbox" class="checkbox">
Edit captions before generation
</label>
</div>
</div>
</div>
`;
$('#extensions_settings2').append(html);
}
addSettings();
addPictureSendForm();
addSendPictureButton();
setImageIcon();
2023-11-07 00:58:34 +01:00
migrateSettings();
2023-07-20 19:32:15 +02:00
moduleWorker();
2023-07-20 19:32:15 +02:00
$('#caption_refine_mode').prop('checked', !!(extension_settings.caption.refine_mode));
$('#caption_source').val(extension_settings.caption.source);
2023-11-07 00:58:34 +01:00
$('#caption_prompt').val(extension_settings.caption.prompt);
$('#caption_template').val(extension_settings.caption.template);
2023-07-20 19:32:15 +02:00
$('#caption_refine_mode').on('input', onRefineModeInput);
$('#caption_source').on('change', () => {
extension_settings.caption.source = String($('#caption_source').val());
saveSettingsDebounced();
});
2023-11-07 00:58:34 +01:00
$('#caption_prompt').on('input', () => {
extension_settings.caption.prompt = String($('#caption_prompt').val());
saveSettingsDebounced();
});
$('#caption_template').on('input', () => {
extension_settings.caption.template = String($('#caption_template').val());
saveSettingsDebounced();
});
2023-07-20 19:32:15 +02:00
setInterval(moduleWorker, UPDATE_INTERVAL);
});