Merge branch 'staging' into neo-server
This commit is contained in:
Cohee
commit
4e1a9da840
|
|
@ -82,6 +82,7 @@ import {
|
|||
flushEphemeralStoppingStrings,
|
||||
context_presets,
|
||||
resetMovableStyles,
|
||||
forceCharacterEditorTokenize,
|
||||
} from './scripts/power-user.js';
|
||||
|
||||
import {
|
||||
|
|
@ -202,7 +203,7 @@ import {
|
|||
selectContextPreset,
|
||||
} from './scripts/instruct-mode.js';
|
||||
import { applyLocale, initLocales } from './scripts/i18n.js';
|
||||
import { getFriendlyTokenizerName, getTokenCount, getTokenizerModel, initTokenizers, saveTokenCache } from './scripts/tokenizers.js';
|
||||
import { getFriendlyTokenizerName, getTokenCount, getTokenCountAsync, getTokenizerModel, initTokenizers, saveTokenCache } from './scripts/tokenizers.js';
|
||||
import { createPersona, initPersonas, selectCurrentPersona, setPersonaDescription, updatePersonaNameIfExists } from './scripts/personas.js';
|
||||
import { getBackgrounds, initBackgrounds, loadBackgroundSettings, background_settings } from './scripts/backgrounds.js';
|
||||
import { hideLoader, showLoader } from './scripts/loader.js';
|
||||
|
|
@ -3469,7 +3470,7 @@ async function Generate(type, { automatic_trigger, force_name2, quiet_prompt, qu
|
|||
let chatString = '';
|
||||
let cyclePrompt = '';
|
||||
|
||||
function getMessagesTokenCount() {
|
||||
async function getMessagesTokenCount() {
|
||||
const encodeString = [
|
||||
beforeScenarioAnchor,
|
||||
storyString,
|
||||
|
|
@ -3480,7 +3481,7 @@ async function Generate(type, { automatic_trigger, force_name2, quiet_prompt, qu
|
|||
cyclePrompt,
|
||||
userAlignmentMessage,
|
||||
].join('').replace(/\r/gm, '');
|
||||
return getTokenCount(encodeString, power_user.token_padding);
|
||||
return getTokenCountAsync(encodeString, power_user.token_padding);
|
||||
}
|
||||
|
||||
// Force pinned examples into the context
|
||||
|
|
@ -3496,7 +3497,7 @@ async function Generate(type, { automatic_trigger, force_name2, quiet_prompt, qu
|
|||
|
||||
// Collect enough messages to fill the context
|
||||
let arrMes = new Array(chat2.length);
|
||||
let tokenCount = getMessagesTokenCount();
|
||||
let tokenCount = await getMessagesTokenCount();
|
||||
let lastAddedIndex = -1;
|
||||
|
||||
// Pre-allocate all injections first.
|
||||
|
|
@ -3508,7 +3509,7 @@ async function Generate(type, { automatic_trigger, force_name2, quiet_prompt, qu
|
|||
continue;
|
||||
}
|
||||
|
||||
tokenCount += getTokenCount(item.replace(/\r/gm, ''));
|
||||
tokenCount += await getTokenCountAsync(item.replace(/\r/gm, ''));
|
||||
chatString = item + chatString;
|
||||
if (tokenCount < this_max_context) {
|
||||
arrMes[index] = item;
|
||||
|
|
@ -3538,7 +3539,7 @@ async function Generate(type, { automatic_trigger, force_name2, quiet_prompt, qu
|
|||
continue;
|
||||
}
|
||||
|
||||
tokenCount += getTokenCount(item.replace(/\r/gm, ''));
|
||||
tokenCount += await getTokenCountAsync(item.replace(/\r/gm, ''));
|
||||
chatString = item + chatString;
|
||||
if (tokenCount < this_max_context) {
|
||||
arrMes[i] = item;
|
||||
|
|
@ -3554,7 +3555,7 @@ async function Generate(type, { automatic_trigger, force_name2, quiet_prompt, qu
|
|||
// Add user alignment message if last message is not a user message
|
||||
const stoppedAtUser = userMessageIndices.includes(lastAddedIndex);
|
||||
if (addUserAlignment && !stoppedAtUser) {
|
||||
tokenCount += getTokenCount(userAlignmentMessage.replace(/\r/gm, ''));
|
||||
tokenCount += await getTokenCountAsync(userAlignmentMessage.replace(/\r/gm, ''));
|
||||
chatString = userAlignmentMessage + chatString;
|
||||
arrMes.push(userAlignmentMessage);
|
||||
injectedIndices.push(arrMes.length - 1);
|
||||
|
|
@ -3580,11 +3581,11 @@ async function Generate(type, { automatic_trigger, force_name2, quiet_prompt, qu
|
|||
}
|
||||
|
||||
// Estimate how many unpinned example messages fit in the context
|
||||
tokenCount = getMessagesTokenCount();
|
||||
tokenCount = await getMessagesTokenCount();
|
||||
let count_exm_add = 0;
|
||||
if (!power_user.pin_examples) {
|
||||
for (let example of mesExamplesArray) {
|
||||
tokenCount += getTokenCount(example.replace(/\r/gm, ''));
|
||||
tokenCount += await getTokenCountAsync(example.replace(/\r/gm, ''));
|
||||
examplesString += example;
|
||||
if (tokenCount < this_max_context) {
|
||||
count_exm_add++;
|
||||
|
|
@ -3739,7 +3740,7 @@ async function Generate(type, { automatic_trigger, force_name2, quiet_prompt, qu
|
|||
return promptCache;
|
||||
}
|
||||
|
||||
function checkPromptSize() {
|
||||
async function checkPromptSize() {
|
||||
console.debug('---checking Prompt size');
|
||||
setPromptString();
|
||||
const prompt = [
|
||||
|
|
@ -3752,15 +3753,15 @@ async function Generate(type, { automatic_trigger, force_name2, quiet_prompt, qu
|
|||
generatedPromptCache,
|
||||
quiet_prompt,
|
||||
].join('').replace(/\r/gm, '');
|
||||
let thisPromptContextSize = getTokenCount(prompt, power_user.token_padding);
|
||||
let thisPromptContextSize = await getTokenCountAsync(prompt, power_user.token_padding);
|
||||
|
||||
if (thisPromptContextSize > this_max_context) { //if the prepared prompt is larger than the max context size...
|
||||
if (count_exm_add > 0) { // ..and we have example mesages..
|
||||
count_exm_add--; // remove the example messages...
|
||||
checkPromptSize(); // and try agin...
|
||||
await checkPromptSize(); // and try agin...
|
||||
} else if (mesSend.length > 0) { // if the chat history is longer than 0
|
||||
mesSend.shift(); // remove the first (oldest) chat entry..
|
||||
checkPromptSize(); // and check size again..
|
||||
await checkPromptSize(); // and check size again..
|
||||
} else {
|
||||
//end
|
||||
console.debug(`---mesSend.length = ${mesSend.length}`);
|
||||
|
|
@ -3770,7 +3771,7 @@ async function Generate(type, { automatic_trigger, force_name2, quiet_prompt, qu
|
|||
|
||||
if (generatedPromptCache.length > 0 && main_api !== 'openai') {
|
||||
console.debug('---Generated Prompt Cache length: ' + generatedPromptCache.length);
|
||||
checkPromptSize();
|
||||
await checkPromptSize();
|
||||
} else {
|
||||
console.debug('---calling setPromptString ' + generatedPromptCache.length);
|
||||
setPromptString();
|
||||
|
|
@ -4433,7 +4434,7 @@ export async function sendMessageAsUser(messageText, messageBias, insertAt = nul
|
|||
};
|
||||
|
||||
if (power_user.message_token_count_enabled) {
|
||||
message.extra.token_count = getTokenCount(message.mes, 0);
|
||||
message.extra.token_count = await getTokenCountAsync(message.mes, 0);
|
||||
}
|
||||
|
||||
// Lock user avatar to a persona.
|
||||
|
|
@ -4596,21 +4597,21 @@ async function promptItemize(itemizedPrompts, requestedMesId) {
|
|||
}
|
||||
|
||||
const params = {
|
||||
charDescriptionTokens: getTokenCount(itemizedPrompts[thisPromptSet].charDescription),
|
||||
charPersonalityTokens: getTokenCount(itemizedPrompts[thisPromptSet].charPersonality),
|
||||
scenarioTextTokens: getTokenCount(itemizedPrompts[thisPromptSet].scenarioText),
|
||||
userPersonaStringTokens: getTokenCount(itemizedPrompts[thisPromptSet].userPersona),
|
||||
worldInfoStringTokens: getTokenCount(itemizedPrompts[thisPromptSet].worldInfoString),
|
||||
allAnchorsTokens: getTokenCount(itemizedPrompts[thisPromptSet].allAnchors),
|
||||
summarizeStringTokens: getTokenCount(itemizedPrompts[thisPromptSet].summarizeString),
|
||||
authorsNoteStringTokens: getTokenCount(itemizedPrompts[thisPromptSet].authorsNoteString),
|
||||
smartContextStringTokens: getTokenCount(itemizedPrompts[thisPromptSet].smartContextString),
|
||||
beforeScenarioAnchorTokens: getTokenCount(itemizedPrompts[thisPromptSet].beforeScenarioAnchor),
|
||||
afterScenarioAnchorTokens: getTokenCount(itemizedPrompts[thisPromptSet].afterScenarioAnchor),
|
||||
zeroDepthAnchorTokens: getTokenCount(itemizedPrompts[thisPromptSet].zeroDepthAnchor), // TODO: unused
|
||||
charDescriptionTokens: await getTokenCountAsync(itemizedPrompts[thisPromptSet].charDescription),
|
||||
charPersonalityTokens: await getTokenCountAsync(itemizedPrompts[thisPromptSet].charPersonality),
|
||||
scenarioTextTokens: await getTokenCountAsync(itemizedPrompts[thisPromptSet].scenarioText),
|
||||
userPersonaStringTokens: await getTokenCountAsync(itemizedPrompts[thisPromptSet].userPersona),
|
||||
worldInfoStringTokens: await getTokenCountAsync(itemizedPrompts[thisPromptSet].worldInfoString),
|
||||
allAnchorsTokens: await getTokenCountAsync(itemizedPrompts[thisPromptSet].allAnchors),
|
||||
summarizeStringTokens: await getTokenCountAsync(itemizedPrompts[thisPromptSet].summarizeString),
|
||||
authorsNoteStringTokens: await getTokenCountAsync(itemizedPrompts[thisPromptSet].authorsNoteString),
|
||||
smartContextStringTokens: await getTokenCountAsync(itemizedPrompts[thisPromptSet].smartContextString),
|
||||
beforeScenarioAnchorTokens: await getTokenCountAsync(itemizedPrompts[thisPromptSet].beforeScenarioAnchor),
|
||||
afterScenarioAnchorTokens: await getTokenCountAsync(itemizedPrompts[thisPromptSet].afterScenarioAnchor),
|
||||
zeroDepthAnchorTokens: await getTokenCountAsync(itemizedPrompts[thisPromptSet].zeroDepthAnchor), // TODO: unused
|
||||
thisPrompt_padding: itemizedPrompts[thisPromptSet].padding,
|
||||
this_main_api: itemizedPrompts[thisPromptSet].main_api,
|
||||
chatInjects: getTokenCount(itemizedPrompts[thisPromptSet].chatInjects),
|
||||
chatInjects: await getTokenCountAsync(itemizedPrompts[thisPromptSet].chatInjects),
|
||||
};
|
||||
|
||||
if (params.chatInjects) {
|
||||
|
|
@ -4664,13 +4665,13 @@ async function promptItemize(itemizedPrompts, requestedMesId) {
|
|||
} else {
|
||||
//for non-OAI APIs
|
||||
//console.log('-- Counting non-OAI Tokens');
|
||||
params.finalPromptTokens = getTokenCount(itemizedPrompts[thisPromptSet].finalPrompt);
|
||||
params.storyStringTokens = getTokenCount(itemizedPrompts[thisPromptSet].storyString) - params.worldInfoStringTokens;
|
||||
params.examplesStringTokens = getTokenCount(itemizedPrompts[thisPromptSet].examplesString);
|
||||
params.mesSendStringTokens = getTokenCount(itemizedPrompts[thisPromptSet].mesSendString);
|
||||
params.finalPromptTokens = await getTokenCountAsync(itemizedPrompts[thisPromptSet].finalPrompt);
|
||||
params.storyStringTokens = await getTokenCountAsync(itemizedPrompts[thisPromptSet].storyString) - params.worldInfoStringTokens;
|
||||
params.examplesStringTokens = await getTokenCountAsync(itemizedPrompts[thisPromptSet].examplesString);
|
||||
params.mesSendStringTokens = await getTokenCountAsync(itemizedPrompts[thisPromptSet].mesSendString);
|
||||
params.ActualChatHistoryTokens = params.mesSendStringTokens - (params.allAnchorsTokens - (params.beforeScenarioAnchorTokens + params.afterScenarioAnchorTokens)) + power_user.token_padding;
|
||||
params.instructionTokens = getTokenCount(itemizedPrompts[thisPromptSet].instruction);
|
||||
params.promptBiasTokens = getTokenCount(itemizedPrompts[thisPromptSet].promptBias);
|
||||
params.instructionTokens = await getTokenCountAsync(itemizedPrompts[thisPromptSet].instruction);
|
||||
params.promptBiasTokens = await getTokenCountAsync(itemizedPrompts[thisPromptSet].promptBias);
|
||||
|
||||
params.totalTokensInPrompt =
|
||||
params.storyStringTokens + //chardefs total
|
||||
|
|
@ -5073,7 +5074,7 @@ async function saveReply(type, getMessage, fromStreaming, title, swipes) {
|
|||
chat[chat.length - 1]['extra']['api'] = getGeneratingApi();
|
||||
chat[chat.length - 1]['extra']['model'] = getGeneratingModel();
|
||||
if (power_user.message_token_count_enabled) {
|
||||
chat[chat.length - 1]['extra']['token_count'] = getTokenCount(chat[chat.length - 1]['mes'], 0);
|
||||
chat[chat.length - 1]['extra']['token_count'] = await getTokenCountAsync(chat[chat.length - 1]['mes'], 0);
|
||||
}
|
||||
const chat_id = (chat.length - 1);
|
||||
await eventSource.emit(event_types.MESSAGE_RECEIVED, chat_id);
|
||||
|
|
@ -5093,7 +5094,7 @@ async function saveReply(type, getMessage, fromStreaming, title, swipes) {
|
|||
chat[chat.length - 1]['extra']['api'] = getGeneratingApi();
|
||||
chat[chat.length - 1]['extra']['model'] = getGeneratingModel();
|
||||
if (power_user.message_token_count_enabled) {
|
||||
chat[chat.length - 1]['extra']['token_count'] = getTokenCount(chat[chat.length - 1]['mes'], 0);
|
||||
chat[chat.length - 1]['extra']['token_count'] = await getTokenCountAsync(chat[chat.length - 1]['mes'], 0);
|
||||
}
|
||||
const chat_id = (chat.length - 1);
|
||||
await eventSource.emit(event_types.MESSAGE_RECEIVED, chat_id);
|
||||
|
|
@ -5110,7 +5111,7 @@ async function saveReply(type, getMessage, fromStreaming, title, swipes) {
|
|||
chat[chat.length - 1]['extra']['api'] = getGeneratingApi();
|
||||
chat[chat.length - 1]['extra']['model'] = getGeneratingModel();
|
||||
if (power_user.message_token_count_enabled) {
|
||||
chat[chat.length - 1]['extra']['token_count'] = getTokenCount(chat[chat.length - 1]['mes'], 0);
|
||||
chat[chat.length - 1]['extra']['token_count'] = await getTokenCountAsync(chat[chat.length - 1]['mes'], 0);
|
||||
}
|
||||
const chat_id = (chat.length - 1);
|
||||
await eventSource.emit(event_types.MESSAGE_RECEIVED, chat_id);
|
||||
|
|
@ -5135,7 +5136,7 @@ async function saveReply(type, getMessage, fromStreaming, title, swipes) {
|
|||
chat[chat.length - 1]['gen_finished'] = generationFinished;
|
||||
|
||||
if (power_user.message_token_count_enabled) {
|
||||
chat[chat.length - 1]['extra']['token_count'] = getTokenCount(chat[chat.length - 1]['mes'], 0);
|
||||
chat[chat.length - 1]['extra']['token_count'] = await getTokenCountAsync(chat[chat.length - 1]['mes'], 0);
|
||||
}
|
||||
|
||||
if (selected_group) {
|
||||
|
|
@ -5841,10 +5842,11 @@ function changeMainAPI() {
|
|||
|
||||
if (main_api == 'koboldhorde') {
|
||||
getStatusHorde();
|
||||
getHordeModels();
|
||||
getHordeModels(true);
|
||||
}
|
||||
|
||||
setupChatCompletionPromptManager(oai_settings);
|
||||
forceCharacterEditorTokenize();
|
||||
}
|
||||
|
||||
////////////////////////////////////////////////////
|
||||
|
|
@ -7854,7 +7856,7 @@ function swipe_left() { // when we swipe left..but no generation.
|
|||
duration: swipe_duration,
|
||||
easing: animation_easing,
|
||||
queue: false,
|
||||
complete: function () {
|
||||
complete: async function () {
|
||||
const is_animation_scroll = ($('#chat').scrollTop() >= ($('#chat').prop('scrollHeight') - $('#chat').outerHeight()) - 10);
|
||||
//console.log('on left swipe click calling addOneMessage');
|
||||
addOneMessage(chat[chat.length - 1], { type: 'swipe' });
|
||||
|
|
@ -7865,7 +7867,7 @@ function swipe_left() { // when we swipe left..but no generation.
|
|||
}
|
||||
|
||||
const swipeMessage = $('#chat').find(`[mesid="${chat.length - 1}"]`);
|
||||
const tokenCount = getTokenCount(chat[chat.length - 1].mes, 0);
|
||||
const tokenCount = await getTokenCountAsync(chat[chat.length - 1].mes, 0);
|
||||
chat[chat.length - 1]['extra']['token_count'] = tokenCount;
|
||||
swipeMessage.find('.tokenCounterDisplay').text(`${tokenCount}t`);
|
||||
}
|
||||
|
|
@ -8030,7 +8032,7 @@ const swipe_right = () => {
|
|||
duration: swipe_duration,
|
||||
easing: animation_easing,
|
||||
queue: false,
|
||||
complete: function () {
|
||||
complete: async function () {
|
||||
/*if (!selected_group) {
|
||||
var typingIndicator = $("#typing_indicator_template .typing_indicator").clone();
|
||||
typingIndicator.find(".typing_indicator_name").text(characters[this_chid].name);
|
||||
|
|
@ -8056,7 +8058,7 @@ const swipe_right = () => {
|
|||
chat[chat.length - 1].extra = {};
|
||||
}
|
||||
|
||||
const tokenCount = getTokenCount(chat[chat.length - 1].mes, 0);
|
||||
const tokenCount = await getTokenCountAsync(chat[chat.length - 1].mes, 0);
|
||||
chat[chat.length - 1]['extra']['token_count'] = tokenCount;
|
||||
swipeMessage.find('.tokenCounterDisplay').text(`${tokenCount}t`);
|
||||
}
|
||||
|
|
@ -8566,7 +8568,7 @@ function addDebugFunctions() {
|
|||
message.extra = {};
|
||||
}
|
||||
|
||||
message.extra.token_count = getTokenCount(message.mes, 0);
|
||||
message.extra.token_count = await getTokenCountAsync(message.mes, 0);
|
||||
}
|
||||
|
||||
await saveChatConditional();
|
||||
|
|
|
|||
|
|
@ -34,7 +34,7 @@ import {
|
|||
} from './secrets.js';
|
||||
import { debounce, delay, getStringHash, isValidUrl } from './utils.js';
|
||||
import { chat_completion_sources, oai_settings } from './openai.js';
|
||||
import { getTokenCount } from './tokenizers.js';
|
||||
import { getTokenCountAsync } from './tokenizers.js';
|
||||
import { textgen_types, textgenerationwebui_settings as textgen_settings, getTextGenServer } from './textgen-settings.js';
|
||||
|
||||
import Bowser from '../lib/bowser.min.js';
|
||||
|
|
@ -51,6 +51,7 @@ var SelectedCharacterTab = document.getElementById('rm_button_selected_ch');
|
|||
|
||||
var connection_made = false;
|
||||
var retry_delay = 500;
|
||||
let counterNonce = Date.now();
|
||||
|
||||
const observerConfig = { childList: true, subtree: true };
|
||||
const countTokensDebounced = debounce(RA_CountCharTokens, 1000);
|
||||
|
|
@ -202,24 +203,32 @@ $('#rm_ch_create_block').on('input', function () { countTokensDebounced(); });
|
|||
//when any input is made to the advanced editing popup textareas
|
||||
$('#character_popup').on('input', function () { countTokensDebounced(); });
|
||||
//function:
|
||||
export function RA_CountCharTokens() {
|
||||
export async function RA_CountCharTokens() {
|
||||
counterNonce = Date.now();
|
||||
const counterNonceLocal = counterNonce;
|
||||
let total_tokens = 0;
|
||||
let permanent_tokens = 0;
|
||||
|
||||
$('[data-token-counter]').each(function () {
|
||||
const counter = $(this);
|
||||
const tokenCounters = document.querySelectorAll('[data-token-counter]');
|
||||
for (const tokenCounter of tokenCounters) {
|
||||
if (counterNonceLocal !== counterNonce) {
|
||||
return;
|
||||
}
|
||||
|
||||
const counter = $(tokenCounter);
|
||||
const input = $(document.getElementById(counter.data('token-counter')));
|
||||
const isPermanent = counter.data('token-permanent') === true;
|
||||
const value = String(input.val());
|
||||
|
||||
if (input.length === 0) {
|
||||
counter.text('Invalid input reference');
|
||||
return;
|
||||
continue;
|
||||
}
|
||||
|
||||
if (!value) {
|
||||
input.data('last-value-hash', '');
|
||||
counter.text(0);
|
||||
return;
|
||||
continue;
|
||||
}
|
||||
|
||||
const valueHash = getStringHash(value);
|
||||
|
|
@ -230,13 +239,18 @@ export function RA_CountCharTokens() {
|
|||
} else {
|
||||
// We substitute macro for existing characters, but not for the character being created
|
||||
const valueToCount = menu_type === 'create' ? value : substituteParams(value);
|
||||
const tokens = getTokenCount(valueToCount);
|
||||
const tokens = await getTokenCountAsync(valueToCount);
|
||||
|
||||
if (counterNonceLocal !== counterNonce) {
|
||||
return;
|
||||
}
|
||||
|
||||
counter.text(tokens);
|
||||
total_tokens += tokens;
|
||||
permanent_tokens += isPermanent ? tokens : 0;
|
||||
input.data('last-value-hash', valueHash);
|
||||
}
|
||||
});
|
||||
}
|
||||
|
||||
// Warn if total tokens exceeds the limit of half the max context
|
||||
const tokenLimit = Math.max(((main_api !== 'openai' ? max_context : oai_settings.openai_max_context) / 2), 1024);
|
||||
|
|
|
|||
|
|
@ -11,7 +11,7 @@ import { selected_group } from './group-chats.js';
|
|||
import { extension_settings, getContext, saveMetadataDebounced } from './extensions.js';
|
||||
import { registerSlashCommand } from './slash-commands.js';
|
||||
import { getCharaFilename, debounce, delay } from './utils.js';
|
||||
import { getTokenCount } from './tokenizers.js';
|
||||
import { getTokenCountAsync } from './tokenizers.js';
|
||||
export { MODULE_NAME as NOTE_MODULE_NAME };
|
||||
|
||||
const MODULE_NAME = '2_floating_prompt'; // <= Deliberate, for sorting lower than memory
|
||||
|
|
@ -84,9 +84,9 @@ function updateSettings() {
|
|||
setFloatingPrompt();
|
||||
}
|
||||
|
||||
const setMainPromptTokenCounterDebounced = debounce((value) => $('#extension_floating_prompt_token_counter').text(getTokenCount(value)), 1000);
|
||||
const setCharaPromptTokenCounterDebounced = debounce((value) => $('#extension_floating_chara_token_counter').text(getTokenCount(value)), 1000);
|
||||
const setDefaultPromptTokenCounterDebounced = debounce((value) => $('#extension_floating_default_token_counter').text(getTokenCount(value)), 1000);
|
||||
const setMainPromptTokenCounterDebounced = debounce(async (value) => $('#extension_floating_prompt_token_counter').text(await getTokenCountAsync(value)), 1000);
|
||||
const setCharaPromptTokenCounterDebounced = debounce(async (value) => $('#extension_floating_chara_token_counter').text(await getTokenCountAsync(value)), 1000);
|
||||
const setDefaultPromptTokenCounterDebounced = debounce(async (value) => $('#extension_floating_default_token_counter').text(await getTokenCountAsync(value)), 1000);
|
||||
|
||||
async function onExtensionFloatingPromptInput() {
|
||||
chat_metadata[metadata_keys.prompt] = $(this).val();
|
||||
|
|
@ -394,7 +394,7 @@ function onANMenuItemClick() {
|
|||
}
|
||||
}
|
||||
|
||||
function onChatChanged() {
|
||||
async function onChatChanged() {
|
||||
loadSettings();
|
||||
setFloatingPrompt();
|
||||
const context = getContext();
|
||||
|
|
@ -402,7 +402,7 @@ function onChatChanged() {
|
|||
// Disable the chara note if in a group
|
||||
$('#extension_floating_chara').prop('disabled', context.groupId ? true : false);
|
||||
|
||||
const tokenCounter1 = chat_metadata[metadata_keys.prompt] ? getTokenCount(chat_metadata[metadata_keys.prompt]) : 0;
|
||||
const tokenCounter1 = chat_metadata[metadata_keys.prompt] ? await getTokenCountAsync(chat_metadata[metadata_keys.prompt]) : 0;
|
||||
$('#extension_floating_prompt_token_counter').text(tokenCounter1);
|
||||
|
||||
let tokenCounter2;
|
||||
|
|
@ -410,15 +410,13 @@ function onChatChanged() {
|
|||
const charaNote = extension_settings.note.chara.find((e) => e.name === getCharaFilename());
|
||||
|
||||
if (charaNote) {
|
||||
tokenCounter2 = getTokenCount(charaNote.prompt);
|
||||
tokenCounter2 = await getTokenCountAsync(charaNote.prompt);
|
||||
}
|
||||
}
|
||||
|
||||
if (tokenCounter2) {
|
||||
$('#extension_floating_chara_token_counter').text(tokenCounter2);
|
||||
}
|
||||
$('#extension_floating_chara_token_counter').text(tokenCounter2 || 0);
|
||||
|
||||
const tokenCounter3 = extension_settings.note.default ? getTokenCount(extension_settings.note.default) : 0;
|
||||
const tokenCounter3 = extension_settings.note.default ? await getTokenCountAsync(extension_settings.note.default) : 0;
|
||||
$('#extension_floating_default_token_counter').text(tokenCounter3);
|
||||
}
|
||||
|
||||
|
|
|
|||
|
|
@ -19,7 +19,7 @@ import { is_group_generating, selected_group } from '../../group-chats.js';
|
|||
import { registerSlashCommand } from '../../slash-commands.js';
|
||||
import { loadMovingUIState } from '../../power-user.js';
|
||||
import { dragElement } from '../../RossAscends-mods.js';
|
||||
import { getTextTokens, getTokenCount, tokenizers } from '../../tokenizers.js';
|
||||
import { getTextTokens, getTokenCountAsync, tokenizers } from '../../tokenizers.js';
|
||||
export { MODULE_NAME };
|
||||
|
||||
const MODULE_NAME = '1_memory';
|
||||
|
|
@ -129,7 +129,7 @@ async function onPromptForceWordsAutoClick() {
|
|||
const allMessages = chat.filter(m => !m.is_system && m.mes).map(m => m.mes);
|
||||
const messagesWordCount = allMessages.map(m => extractAllWords(m)).flat().length;
|
||||
const averageMessageWordCount = messagesWordCount / allMessages.length;
|
||||
const tokensPerWord = getTokenCount(allMessages.join('\n')) / messagesWordCount;
|
||||
const tokensPerWord = await getTokenCountAsync(allMessages.join('\n')) / messagesWordCount;
|
||||
const wordsPerToken = 1 / tokensPerWord;
|
||||
const maxPromptLengthWords = Math.round(maxPromptLength * wordsPerToken);
|
||||
// How many words should pass so that messages will start be dropped out of context;
|
||||
|
|
@ -166,11 +166,11 @@ async function onPromptIntervalAutoClick() {
|
|||
const chat = context.chat;
|
||||
const allMessages = chat.filter(m => !m.is_system && m.mes).map(m => m.mes);
|
||||
const messagesWordCount = allMessages.map(m => extractAllWords(m)).flat().length;
|
||||
const messagesTokenCount = getTokenCount(allMessages.join('\n'));
|
||||
const messagesTokenCount = await getTokenCountAsync(allMessages.join('\n'));
|
||||
const tokensPerWord = messagesTokenCount / messagesWordCount;
|
||||
const averageMessageTokenCount = messagesTokenCount / allMessages.length;
|
||||
const targetSummaryTokens = Math.round(extension_settings.memory.promptWords * tokensPerWord);
|
||||
const promptTokens = getTokenCount(extension_settings.memory.prompt);
|
||||
const promptTokens = await getTokenCountAsync(extension_settings.memory.prompt);
|
||||
const promptAllowance = maxPromptLength - promptTokens - targetSummaryTokens;
|
||||
const maxMessagesPerSummary = extension_settings.memory.maxMessagesPerRequest || 0;
|
||||
const averageMessagesPerPrompt = Math.floor(promptAllowance / averageMessageTokenCount);
|
||||
|
|
@ -603,8 +603,7 @@ async function getRawSummaryPrompt(context, prompt) {
|
|||
const entry = `${message.name}:\n${message.mes}`;
|
||||
chatBuffer.push(entry);
|
||||
|
||||
const tokens = getTokenCount(getMemoryString(true), PADDING);
|
||||
await delay(1);
|
||||
const tokens = await getTokenCountAsync(getMemoryString(true), PADDING);
|
||||
|
||||
if (tokens > PROMPT_SIZE) {
|
||||
chatBuffer.pop();
|
||||
|
|
|
|||
|
|
@ -1,7 +1,7 @@
|
|||
import { callPopup, main_api } from '../../../script.js';
|
||||
import { getContext } from '../../extensions.js';
|
||||
import { registerSlashCommand } from '../../slash-commands.js';
|
||||
import { getFriendlyTokenizerName, getTextTokens, getTokenCount, tokenizers } from '../../tokenizers.js';
|
||||
import { getFriendlyTokenizerName, getTextTokens, getTokenCountAsync, tokenizers } from '../../tokenizers.js';
|
||||
import { resetScrollHeight, debounce } from '../../utils.js';
|
||||
|
||||
function rgb2hex(rgb) {
|
||||
|
|
@ -38,7 +38,7 @@ async function doTokenCounter() {
|
|||
</div>`;
|
||||
|
||||
const dialog = $(html);
|
||||
const countDebounced = debounce(() => {
|
||||
const countDebounced = debounce(async () => {
|
||||
const text = String($('#token_counter_textarea').val());
|
||||
const ids = main_api == 'openai' ? getTextTokens(tokenizers.OPENAI, text) : getTextTokens(tokenizerId, text);
|
||||
|
||||
|
|
@ -50,8 +50,7 @@ async function doTokenCounter() {
|
|||
drawChunks(Object.getOwnPropertyDescriptor(ids, 'chunks').value, ids);
|
||||
}
|
||||
} else {
|
||||
const context = getContext();
|
||||
const count = context.getTokenCount(text);
|
||||
const count = await getTokenCountAsync(text);
|
||||
$('#token_counter_ids').text('—');
|
||||
$('#token_counter_result').text(count);
|
||||
$('#tokenized_chunks_display').text('—');
|
||||
|
|
@ -109,7 +108,7 @@ function drawChunks(chunks, ids) {
|
|||
}
|
||||
}
|
||||
|
||||
function doCount() {
|
||||
async function doCount() {
|
||||
// get all of the messages in the chat
|
||||
const context = getContext();
|
||||
const messages = context.chat.filter(x => x.mes && !x.is_system).map(x => x.mes);
|
||||
|
|
@ -120,7 +119,8 @@ function doCount() {
|
|||
console.debug('All messages:', allMessages);
|
||||
|
||||
//toastr success with the token count of the chat
|
||||
toastr.success(`Token count: ${getTokenCount(allMessages)}`);
|
||||
const count = await getTokenCountAsync(allMessages);
|
||||
toastr.success(`Token count: ${count}`);
|
||||
}
|
||||
|
||||
jQuery(() => {
|
||||
|
|
|
|||
|
|
@ -42,7 +42,7 @@ import {
|
|||
promptManagerDefaultPromptOrders,
|
||||
} from './PromptManager.js';
|
||||
|
||||
import { getCustomStoppingStrings, persona_description_positions, power_user } from './power-user.js';
|
||||
import { forceCharacterEditorTokenize, getCustomStoppingStrings, persona_description_positions, power_user } from './power-user.js';
|
||||
import { SECRET_KEYS, secret_state, writeSecret } from './secrets.js';
|
||||
|
||||
import { getEventSourceStream } from './sse-stream.js';
|
||||
|
|
@ -3566,7 +3566,7 @@ async function onModelChange() {
|
|||
|
||||
if (oai_settings.chat_completion_source == chat_completion_sources.MAKERSUITE) {
|
||||
if (oai_settings.max_context_unlocked) {
|
||||
$('#openai_max_context').attr('max', unlocked_max);
|
||||
$('#openai_max_context').attr('max', max_1mil);
|
||||
} else if (value === 'gemini-1.5-pro-latest') {
|
||||
$('#openai_max_context').attr('max', max_1mil);
|
||||
} else if (value === 'gemini-ultra' || value === 'gemini-1.0-pro-latest' || value === 'gemini-pro' || value === 'gemini-1.0-ultra-latest') {
|
||||
|
|
@ -4429,6 +4429,7 @@ $(document).ready(async function () {
|
|||
toggleChatCompletionForms();
|
||||
saveSettingsDebounced();
|
||||
reconnectOpenAi();
|
||||
forceCharacterEditorTokenize();
|
||||
eventSource.emit(event_types.CHATCOMPLETION_SOURCE_CHANGED, oai_settings.chat_completion_source);
|
||||
});
|
||||
|
||||
|
|
|
|||
|
|
@ -17,7 +17,7 @@ import {
|
|||
user_avatar,
|
||||
} from '../script.js';
|
||||
import { persona_description_positions, power_user } from './power-user.js';
|
||||
import { getTokenCount } from './tokenizers.js';
|
||||
import { getTokenCountAsync } from './tokenizers.js';
|
||||
import { debounce, delay, download, parseJsonFile } from './utils.js';
|
||||
|
||||
const GRID_STORAGE_KEY = 'Personas_GridView';
|
||||
|
|
@ -171,9 +171,9 @@ export async function convertCharacterToPersona(characterId = null) {
|
|||
/**
|
||||
* Counts the number of tokens in a persona description.
|
||||
*/
|
||||
const countPersonaDescriptionTokens = debounce(() => {
|
||||
const countPersonaDescriptionTokens = debounce(async () => {
|
||||
const description = String($('#persona_description').val());
|
||||
const count = getTokenCount(description);
|
||||
const count = await getTokenCountAsync(description);
|
||||
$('#persona_description_token_count').text(String(count));
|
||||
}, 1000);
|
||||
|
||||
|
|
|
|||
|
|
@ -2764,6 +2764,14 @@ export function getCustomStoppingStrings(limit = undefined) {
|
|||
return strings;
|
||||
}
|
||||
|
||||
export function forceCharacterEditorTokenize() {
|
||||
$('[data-token-counter]').each(function () {
|
||||
$(document.getElementById($(this).data('token-counter'))).data('last-value-hash', '');
|
||||
});
|
||||
$('#rm_ch_create_block').trigger('input');
|
||||
$('#character_popup').trigger('input');
|
||||
}
|
||||
|
||||
$(document).ready(() => {
|
||||
const adjustAutocompleteDebounced = debounce(() => {
|
||||
$('.ui-autocomplete-input').each(function () {
|
||||
|
|
@ -3175,8 +3183,7 @@ $(document).ready(() => {
|
|||
saveSettingsDebounced();
|
||||
|
||||
// Trigger character editor re-tokenize
|
||||
$('#rm_ch_create_block').trigger('input');
|
||||
$('#character_popup').trigger('input');
|
||||
forceCharacterEditorTokenize();
|
||||
});
|
||||
|
||||
$('#send_on_enter').on('change', function () {
|
||||
|
|
|
|||
|
|
@ -46,7 +46,7 @@ import { chat_completion_sources, oai_settings } from './openai.js';
|
|||
import { autoSelectPersona } from './personas.js';
|
||||
import { addEphemeralStoppingString, chat_styles, flushEphemeralStoppingStrings, power_user } from './power-user.js';
|
||||
import { textgen_types, textgenerationwebui_settings } from './textgen-settings.js';
|
||||
import { decodeTextTokens, getFriendlyTokenizerName, getTextTokens, getTokenCount } from './tokenizers.js';
|
||||
import { decodeTextTokens, getFriendlyTokenizerName, getTextTokens, getTokenCountAsync } from './tokenizers.js';
|
||||
import { delay, isFalseBoolean, isTrueBoolean, stringToRange, trimToEndSentence, trimToStartSentence, waitUntilCondition } from './utils.js';
|
||||
import { registerVariableCommands, resolveVariable } from './variables.js';
|
||||
import { background_settings } from './backgrounds.js';
|
||||
|
|
@ -249,7 +249,7 @@ parser.addCommand('trimend', trimEndCallback, [], '<span class="monospace">(text
|
|||
parser.addCommand('inject', injectCallback, [], '<span class="monospace">id=injectId (position=before/after/chat depth=number scan=true/false role=system/user/assistant [text])</span> – injects a text into the LLM prompt for the current chat. Requires a unique injection ID. Positions: "before" main prompt, "after" main prompt, in-"chat" (default: after). Depth: injection depth for the prompt (default: 4). Role: role for in-chat injections (default: system). Scan: include injection content into World Info scans (default: false).', true, true);
|
||||
parser.addCommand('listinjects', listInjectsCallback, [], ' – lists all script injections for the current chat.', true, true);
|
||||
parser.addCommand('flushinjects', flushInjectsCallback, [], ' – removes all script injections for the current chat.', true, true);
|
||||
parser.addCommand('tokens', (_, text) => getTokenCount(text), [], '<span class="monospace">(text)</span> – counts the number of tokens in the text.', true, true);
|
||||
parser.addCommand('tokens', (_, text) => getTokenCountAsync(text), [], '<span class="monospace">(text)</span> – counts the number of tokens in the text.', true, true);
|
||||
parser.addCommand('model', modelCallback, [], '<span class="monospace">(model name)</span> – sets the model for the current API. Gets the current model name if no argument is provided.', true, true);
|
||||
registerVariableCommands();
|
||||
|
||||
|
|
@ -388,7 +388,7 @@ function trimEndCallback(_, value) {
|
|||
return trimToEndSentence(value);
|
||||
}
|
||||
|
||||
function trimTokensCallback(arg, value) {
|
||||
async function trimTokensCallback(arg, value) {
|
||||
if (!value) {
|
||||
console.warn('WARN: No argument provided for /trimtokens command');
|
||||
return '';
|
||||
|
|
@ -406,7 +406,7 @@ function trimTokensCallback(arg, value) {
|
|||
}
|
||||
|
||||
const direction = arg.direction || 'end';
|
||||
const tokenCount = getTokenCount(value);
|
||||
const tokenCount = await getTokenCountAsync(value);
|
||||
|
||||
// Token count is less than the limit, do nothing
|
||||
if (tokenCount <= limit) {
|
||||
|
|
|
|||
|
|
@ -256,11 +256,93 @@ function callTokenizer(type, str) {
|
|||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Calls the underlying tokenizer model to the token count for a string.
|
||||
* @param {number} type Tokenizer type.
|
||||
* @param {string} str String to tokenize.
|
||||
* @returns {Promise<number>} Token count.
|
||||
*/
|
||||
function callTokenizerAsync(type, str) {
|
||||
return new Promise(resolve => {
|
||||
if (type === tokenizers.NONE) {
|
||||
return resolve(guesstimate(str));
|
||||
}
|
||||
|
||||
switch (type) {
|
||||
case tokenizers.API_CURRENT:
|
||||
return callTokenizerAsync(currentRemoteTokenizerAPI(), str).then(resolve);
|
||||
case tokenizers.API_KOBOLD:
|
||||
return countTokensFromKoboldAPI(str, resolve);
|
||||
case tokenizers.API_TEXTGENERATIONWEBUI:
|
||||
return countTokensFromTextgenAPI(str, resolve);
|
||||
default: {
|
||||
const endpointUrl = TOKENIZER_URLS[type]?.count;
|
||||
if (!endpointUrl) {
|
||||
console.warn('Unknown tokenizer type', type);
|
||||
return resolve(apiFailureTokenCount(str));
|
||||
}
|
||||
return countTokensFromServer(endpointUrl, str, resolve);
|
||||
}
|
||||
}
|
||||
});
|
||||
}
|
||||
|
||||
/**
|
||||
* Gets the token count for a string using the current model tokenizer.
|
||||
* @param {string} str String to tokenize
|
||||
* @param {number | undefined} padding Optional padding tokens. Defaults to 0.
|
||||
* @returns {Promise<number>} Token count.
|
||||
*/
|
||||
export async function getTokenCountAsync(str, padding = undefined) {
|
||||
if (typeof str !== 'string' || !str?.length) {
|
||||
return 0;
|
||||
}
|
||||
|
||||
let tokenizerType = power_user.tokenizer;
|
||||
|
||||
if (main_api === 'openai') {
|
||||
if (padding === power_user.token_padding) {
|
||||
// For main "shadow" prompt building
|
||||
tokenizerType = tokenizers.NONE;
|
||||
} else {
|
||||
// For extensions and WI
|
||||
return counterWrapperOpenAIAsync(str);
|
||||
}
|
||||
}
|
||||
|
||||
if (tokenizerType === tokenizers.BEST_MATCH) {
|
||||
tokenizerType = getTokenizerBestMatch(main_api);
|
||||
}
|
||||
|
||||
if (padding === undefined) {
|
||||
padding = 0;
|
||||
}
|
||||
|
||||
const cacheObject = getTokenCacheObject();
|
||||
const hash = getStringHash(str);
|
||||
const cacheKey = `${tokenizerType}-${hash}+${padding}`;
|
||||
|
||||
if (typeof cacheObject[cacheKey] === 'number') {
|
||||
return cacheObject[cacheKey];
|
||||
}
|
||||
|
||||
const result = (await callTokenizerAsync(tokenizerType, str)) + padding;
|
||||
|
||||
if (isNaN(result)) {
|
||||
console.warn('Token count calculation returned NaN');
|
||||
return 0;
|
||||
}
|
||||
|
||||
cacheObject[cacheKey] = result;
|
||||
return result;
|
||||
}
|
||||
|
||||
/**
|
||||
* Gets the token count for a string using the current model tokenizer.
|
||||
* @param {string} str String to tokenize
|
||||
* @param {number | undefined} padding Optional padding tokens. Defaults to 0.
|
||||
* @returns {number} Token count.
|
||||
* @deprecated Use getTokenCountAsync instead.
|
||||
*/
|
||||
export function getTokenCount(str, padding = undefined) {
|
||||
if (typeof str !== 'string' || !str?.length) {
|
||||
|
|
@ -310,12 +392,23 @@ export function getTokenCount(str, padding = undefined) {
|
|||
* Gets the token count for a string using the OpenAI tokenizer.
|
||||
* @param {string} text Text to tokenize.
|
||||
* @returns {number} Token count.
|
||||
* @deprecated Use counterWrapperOpenAIAsync instead.
|
||||
*/
|
||||
function counterWrapperOpenAI(text) {
|
||||
const message = { role: 'system', content: text };
|
||||
return countTokensOpenAI(message, true);
|
||||
}
|
||||
|
||||
/**
|
||||
* Gets the token count for a string using the OpenAI tokenizer.
|
||||
* @param {string} text Text to tokenize.
|
||||
* @returns {Promise<number>} Token count.
|
||||
*/
|
||||
function counterWrapperOpenAIAsync(text) {
|
||||
const message = { role: 'system', content: text };
|
||||
return countTokensOpenAIAsync(message, true);
|
||||
}
|
||||
|
||||
export function getTokenizerModel() {
|
||||
// OpenAI models always provide their own tokenizer
|
||||
if (oai_settings.chat_completion_source == chat_completion_sources.OPENAI) {
|
||||
|
|
@ -410,6 +503,7 @@ export function getTokenizerModel() {
|
|||
|
||||
/**
|
||||
* @param {any[] | Object} messages
|
||||
* @deprecated Use countTokensOpenAIAsync instead.
|
||||
*/
|
||||
export function countTokensOpenAI(messages, full = false) {
|
||||
const shouldTokenizeAI21 = oai_settings.chat_completion_source === chat_completion_sources.AI21 && oai_settings.use_ai21_tokenizer;
|
||||
|
|
@ -466,6 +560,66 @@ export function countTokensOpenAI(messages, full = false) {
|
|||
return token_count;
|
||||
}
|
||||
|
||||
/**
|
||||
* Returns the token count for a message using the OpenAI tokenizer.
|
||||
* @param {object[]|object} messages
|
||||
* @param {boolean} full
|
||||
* @returns {Promise<number>} Token count.
|
||||
*/
|
||||
export async function countTokensOpenAIAsync(messages, full = false) {
|
||||
const shouldTokenizeAI21 = oai_settings.chat_completion_source === chat_completion_sources.AI21 && oai_settings.use_ai21_tokenizer;
|
||||
const shouldTokenizeGoogle = oai_settings.chat_completion_source === chat_completion_sources.MAKERSUITE && oai_settings.use_google_tokenizer;
|
||||
let tokenizerEndpoint = '';
|
||||
if (shouldTokenizeAI21) {
|
||||
tokenizerEndpoint = '/api/tokenizers/ai21/count';
|
||||
} else if (shouldTokenizeGoogle) {
|
||||
tokenizerEndpoint = `/api/tokenizers/google/count?model=${getTokenizerModel()}`;
|
||||
} else {
|
||||
tokenizerEndpoint = `/api/tokenizers/openai/count?model=${getTokenizerModel()}`;
|
||||
}
|
||||
const cacheObject = getTokenCacheObject();
|
||||
|
||||
if (!Array.isArray(messages)) {
|
||||
messages = [messages];
|
||||
}
|
||||
|
||||
let token_count = -1;
|
||||
|
||||
for (const message of messages) {
|
||||
const model = getTokenizerModel();
|
||||
|
||||
if (model === 'claude' || shouldTokenizeAI21 || shouldTokenizeGoogle) {
|
||||
full = true;
|
||||
}
|
||||
|
||||
const hash = getStringHash(JSON.stringify(message));
|
||||
const cacheKey = `${model}-${hash}`;
|
||||
const cachedCount = cacheObject[cacheKey];
|
||||
|
||||
if (typeof cachedCount === 'number') {
|
||||
token_count += cachedCount;
|
||||
}
|
||||
|
||||
else {
|
||||
const data = await jQuery.ajax({
|
||||
async: true,
|
||||
type: 'POST', //
|
||||
url: tokenizerEndpoint,
|
||||
data: JSON.stringify([message]),
|
||||
dataType: 'json',
|
||||
contentType: 'application/json',
|
||||
});
|
||||
|
||||
token_count += Number(data.token_count);
|
||||
cacheObject[cacheKey] = Number(data.token_count);
|
||||
}
|
||||
}
|
||||
|
||||
if (!full) token_count -= 2;
|
||||
|
||||
return token_count;
|
||||
}
|
||||
|
||||
/**
|
||||
* Gets the token cache object for the current chat.
|
||||
* @returns {Object} Token cache object for the current chat.
|
||||
|
|
@ -495,13 +649,15 @@ function getTokenCacheObject() {
|
|||
* Count tokens using the server API.
|
||||
* @param {string} endpoint API endpoint.
|
||||
* @param {string} str String to tokenize.
|
||||
* @param {function} [resolve] Promise resolve function.s
|
||||
* @returns {number} Token count.
|
||||
*/
|
||||
function countTokensFromServer(endpoint, str) {
|
||||
function countTokensFromServer(endpoint, str, resolve) {
|
||||
const isAsync = typeof resolve === 'function';
|
||||
let tokenCount = 0;
|
||||
|
||||
jQuery.ajax({
|
||||
async: false,
|
||||
async: isAsync,
|
||||
type: 'POST',
|
||||
url: endpoint,
|
||||
data: JSON.stringify({ text: str }),
|
||||
|
|
@ -513,6 +669,8 @@ function countTokensFromServer(endpoint, str) {
|
|||
} else {
|
||||
tokenCount = apiFailureTokenCount(str);
|
||||
}
|
||||
|
||||
isAsync && resolve(tokenCount);
|
||||
},
|
||||
});
|
||||
|
||||
|
|
@ -522,13 +680,15 @@ function countTokensFromServer(endpoint, str) {
|
|||
/**
|
||||
* Count tokens using the AI provider's API.
|
||||
* @param {string} str String to tokenize.
|
||||
* @param {function} [resolve] Promise resolve function.
|
||||
* @returns {number} Token count.
|
||||
*/
|
||||
function countTokensFromKoboldAPI(str) {
|
||||
function countTokensFromKoboldAPI(str, resolve) {
|
||||
const isAsync = typeof resolve === 'function';
|
||||
let tokenCount = 0;
|
||||
|
||||
jQuery.ajax({
|
||||
async: false,
|
||||
async: isAsync,
|
||||
type: 'POST',
|
||||
url: TOKENIZER_URLS[tokenizers.API_KOBOLD].count,
|
||||
data: JSON.stringify({
|
||||
|
|
@ -543,6 +703,8 @@ function countTokensFromKoboldAPI(str) {
|
|||
} else {
|
||||
tokenCount = apiFailureTokenCount(str);
|
||||
}
|
||||
|
||||
isAsync && resolve(tokenCount);
|
||||
},
|
||||
});
|
||||
|
||||
|
|
@ -561,13 +723,15 @@ function getTextgenAPITokenizationParams(str) {
|
|||
/**
|
||||
* Count tokens using the AI provider's API.
|
||||
* @param {string} str String to tokenize.
|
||||
* @param {function} [resolve] Promise resolve function.
|
||||
* @returns {number} Token count.
|
||||
*/
|
||||
function countTokensFromTextgenAPI(str) {
|
||||
function countTokensFromTextgenAPI(str, resolve) {
|
||||
const isAsync = typeof resolve === 'function';
|
||||
let tokenCount = 0;
|
||||
|
||||
jQuery.ajax({
|
||||
async: false,
|
||||
async: isAsync,
|
||||
type: 'POST',
|
||||
url: TOKENIZER_URLS[tokenizers.API_TEXTGENERATIONWEBUI].count,
|
||||
data: JSON.stringify(getTextgenAPITokenizationParams(str)),
|
||||
|
|
@ -579,6 +743,8 @@ function countTokensFromTextgenAPI(str) {
|
|||
} else {
|
||||
tokenCount = apiFailureTokenCount(str);
|
||||
}
|
||||
|
||||
isAsync && resolve(tokenCount);
|
||||
},
|
||||
});
|
||||
|
||||
|
|
@ -605,12 +771,14 @@ function apiFailureTokenCount(str) {
|
|||
* Calls the underlying tokenizer model to encode a string to tokens.
|
||||
* @param {string} endpoint API endpoint.
|
||||
* @param {string} str String to tokenize.
|
||||
* @param {function} [resolve] Promise resolve function.
|
||||
* @returns {number[]} Array of token ids.
|
||||
*/
|
||||
function getTextTokensFromServer(endpoint, str) {
|
||||
function getTextTokensFromServer(endpoint, str, resolve) {
|
||||
const isAsync = typeof resolve === 'function';
|
||||
let ids = [];
|
||||
jQuery.ajax({
|
||||
async: false,
|
||||
async: isAsync,
|
||||
type: 'POST',
|
||||
url: endpoint,
|
||||
data: JSON.stringify({ text: str }),
|
||||
|
|
@ -623,6 +791,8 @@ function getTextTokensFromServer(endpoint, str) {
|
|||
if (Array.isArray(data.chunks)) {
|
||||
Object.defineProperty(ids, 'chunks', { value: data.chunks });
|
||||
}
|
||||
|
||||
isAsync && resolve(ids);
|
||||
},
|
||||
});
|
||||
return ids;
|
||||
|
|
@ -631,12 +801,14 @@ function getTextTokensFromServer(endpoint, str) {
|
|||
/**
|
||||
* Calls the AI provider's tokenize API to encode a string to tokens.
|
||||
* @param {string} str String to tokenize.
|
||||
* @param {function} [resolve] Promise resolve function.
|
||||
* @returns {number[]} Array of token ids.
|
||||
*/
|
||||
function getTextTokensFromTextgenAPI(str) {
|
||||
function getTextTokensFromTextgenAPI(str, resolve) {
|
||||
const isAsync = typeof resolve === 'function';
|
||||
let ids = [];
|
||||
jQuery.ajax({
|
||||
async: false,
|
||||
async: isAsync,
|
||||
type: 'POST',
|
||||
url: TOKENIZER_URLS[tokenizers.API_TEXTGENERATIONWEBUI].encode,
|
||||
data: JSON.stringify(getTextgenAPITokenizationParams(str)),
|
||||
|
|
@ -644,6 +816,7 @@ function getTextTokensFromTextgenAPI(str) {
|
|||
contentType: 'application/json',
|
||||
success: function (data) {
|
||||
ids = data.ids;
|
||||
isAsync && resolve(ids);
|
||||
},
|
||||
});
|
||||
return ids;
|
||||
|
|
@ -652,13 +825,15 @@ function getTextTokensFromTextgenAPI(str) {
|
|||
/**
|
||||
* Calls the AI provider's tokenize API to encode a string to tokens.
|
||||
* @param {string} str String to tokenize.
|
||||
* @param {function} [resolve] Promise resolve function.
|
||||
* @returns {number[]} Array of token ids.
|
||||
*/
|
||||
function getTextTokensFromKoboldAPI(str) {
|
||||
function getTextTokensFromKoboldAPI(str, resolve) {
|
||||
const isAsync = typeof resolve === 'function';
|
||||
let ids = [];
|
||||
|
||||
jQuery.ajax({
|
||||
async: false,
|
||||
async: isAsync,
|
||||
type: 'POST',
|
||||
url: TOKENIZER_URLS[tokenizers.API_KOBOLD].encode,
|
||||
data: JSON.stringify({
|
||||
|
|
@ -669,6 +844,7 @@ function getTextTokensFromKoboldAPI(str) {
|
|||
contentType: 'application/json',
|
||||
success: function (data) {
|
||||
ids = data.ids;
|
||||
isAsync && resolve(ids);
|
||||
},
|
||||
});
|
||||
|
||||
|
|
@ -679,13 +855,15 @@ function getTextTokensFromKoboldAPI(str) {
|
|||
* Calls the underlying tokenizer model to decode token ids to text.
|
||||
* @param {string} endpoint API endpoint.
|
||||
* @param {number[]} ids Array of token ids
|
||||
* @param {function} [resolve] Promise resolve function.
|
||||
* @returns {({ text: string, chunks?: string[] })} Decoded token text as a single string and individual chunks (if available).
|
||||
*/
|
||||
function decodeTextTokensFromServer(endpoint, ids) {
|
||||
function decodeTextTokensFromServer(endpoint, ids, resolve) {
|
||||
const isAsync = typeof resolve === 'function';
|
||||
let text = '';
|
||||
let chunks = [];
|
||||
jQuery.ajax({
|
||||
async: false,
|
||||
async: isAsync,
|
||||
type: 'POST',
|
||||
url: endpoint,
|
||||
data: JSON.stringify({ ids: ids }),
|
||||
|
|
@ -694,6 +872,7 @@ function decodeTextTokensFromServer(endpoint, ids) {
|
|||
success: function (data) {
|
||||
text = data.text;
|
||||
chunks = data.chunks;
|
||||
isAsync && resolve({ text, chunks });
|
||||
},
|
||||
});
|
||||
return { text, chunks };
|
||||
|
|
|
|||
|
|
@ -5,7 +5,7 @@ import { NOTE_MODULE_NAME, metadata_keys, shouldWIAddPrompt } from './authors-no
|
|||
import { registerSlashCommand } from './slash-commands.js';
|
||||
import { isMobile } from './RossAscends-mods.js';
|
||||
import { FILTER_TYPES, FilterHelper } from './filters.js';
|
||||
import { getTokenCount } from './tokenizers.js';
|
||||
import { getTokenCountAsync } from './tokenizers.js';
|
||||
import { power_user } from './power-user.js';
|
||||
import { getTagKeyForEntity } from './tags.js';
|
||||
import { resolveVariable } from './variables.js';
|
||||
|
|
@ -1189,8 +1189,8 @@ function getWorldEntry(name, data, entry) {
|
|||
|
||||
// content
|
||||
const counter = template.find('.world_entry_form_token_counter');
|
||||
const countTokensDebounced = debounce(function (counter, value) {
|
||||
const numberOfTokens = getTokenCount(value);
|
||||
const countTokensDebounced = debounce(async function (counter, value) {
|
||||
const numberOfTokens = await getTokenCountAsync(value);
|
||||
$(counter).text(numberOfTokens);
|
||||
}, 1000);
|
||||
|
||||
|
|
@ -2177,7 +2177,7 @@ async function checkWorldInfo(chat, maxContext) {
|
|||
const newEntries = [...activatedNow]
|
||||
.sort((a, b) => sortedEntries.indexOf(a) - sortedEntries.indexOf(b));
|
||||
let newContent = '';
|
||||
const textToScanTokens = getTokenCount(allActivatedText);
|
||||
const textToScanTokens = await getTokenCountAsync(allActivatedText);
|
||||
const probabilityChecksBefore = failedProbabilityChecks.size;
|
||||
|
||||
filterByInclusionGroups(newEntries, allActivatedEntries);
|
||||
|
|
@ -2194,7 +2194,7 @@ async function checkWorldInfo(chat, maxContext) {
|
|||
|
||||
newContent += `${substituteParams(entry.content)}\n`;
|
||||
|
||||
if (textToScanTokens + getTokenCount(newContent) >= budget) {
|
||||
if ((textToScanTokens + (await getTokenCountAsync(newContent))) >= budget) {
|
||||
console.debug('WI budget reached, stopping');
|
||||
if (world_info_overflow_alert) {
|
||||
console.log('Alerting');
|
||||
|
|
|
|||
17
server.js
17
server.js
|
|
@ -556,12 +556,14 @@ const setupTasks = async function () {
|
|||
await statsEndpoint.init();
|
||||
|
||||
const cleanupPlugins = await loadPlugins();
|
||||
const consoleTitle = process.title;
|
||||
|
||||
const exitProcess = async () => {
|
||||
statsEndpoint.onExit();
|
||||
if (typeof cleanupPlugins === 'function') {
|
||||
await cleanupPlugins();
|
||||
}
|
||||
setWindowTitle(consoleTitle);
|
||||
process.exit();
|
||||
};
|
||||
|
||||
|
|
@ -578,6 +580,8 @@ const setupTasks = async function () {
|
|||
|
||||
if (autorun) open(autorunUrl.toString());
|
||||
|
||||
setWindowTitle('SillyTavern WebServer');
|
||||
|
||||
console.log(color.green('SillyTavern is listening on: ' + tavernUrl));
|
||||
|
||||
if (listen) {
|
||||
|
|
@ -619,6 +623,19 @@ if (listen && !enableWhitelist && !basicAuthMode) {
|
|||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Set the title of the terminal window
|
||||
* @param {string} title Desired title for the window
|
||||
*/
|
||||
function setWindowTitle(title) {
|
||||
if (process.platform === 'win32') {
|
||||
process.title = title;
|
||||
}
|
||||
else {
|
||||
process.stdout.write(`\x1b]2;${title}\x1b\x5c`);
|
||||
}
|
||||
}
|
||||
|
||||
if (cliArguments.ssl) {
|
||||
https.createServer(
|
||||
{
|
||||
|
|
|
|||
Loading…
Reference in New Issue