Merge 305afb3713 into be7eb8b2b5

style: `.mes_block { overflow-y: clip; }`

public/script.js

@@ -10516,6 +10516,7 @@ jQuery(async function () {
             <li>Chub Lorebook (Direct Link or ID)<br>Example: <tt>lorebooks/bartleby/example-lorebook</tt></li>
             <li>JanitorAI Character (Direct Link or UUID)<br>Example: <tt>ddd1498a-a370-4136-b138-a8cd9461fdfe_character-aqua-the-useless-goddess</tt></li>
             <li>Pygmalion.chat Character (Direct Link or UUID)<br>Example: <tt>a7ca95a1-0c88-4e23-91b3-149db1e78ab9</tt></li>
+            <li>AICharacterCard.com Character (Direct Link or ID)<br>Example: <tt>AICC/aicharcards/the-game-master</tt></li>
             <li>More coming soon...</li>
         <ul>`;
         const input = await callPopup(html, 'input', '', { okButton: 'Import', rows: 4 });

public/scripts/openai.js

@@ -3250,7 +3250,8 @@ async function onExportPresetClick() {
     delete preset.proxy_password;
 
     const presetJsonString = JSON.stringify(preset, null, 4);
-    download(presetJsonString, oai_settings.preset_settings_openai, 'application/json');
+    const presetFileName = `${oai_settings.preset_settings_openai}.json`;
+    download(presetJsonString, presetFileName, 'application/json');
 }
 
 async function onLogitBiasPresetImportFileChange(e) {
@@ -3298,7 +3299,8 @@ function onLogitBiasPresetExportClick() {
     }
 
     const presetJsonString = JSON.stringify(oai_settings.bias_presets[oai_settings.bias_preset_selected], null, 4);
-    download(presetJsonString, oai_settings.bias_preset_selected, 'application/json');
+    const presetFileName = `${oai_settings.bias_preset_selected}.json`;
+    download(presetJsonString, presetFileName, 'application/json');
 }
 
 async function onDeletePresetClick() {

public/style.css

@@ -1000,6 +1000,7 @@ body .panelControlBar {
     padding-left: 10px;
     width: 100%;
     overflow-x: hidden;
+    overflow-y: clip;
 }
 
 .mes_text {

server.js

@@ -45,7 +45,6 @@ const {
     forwardFetchResponse,
 } = require('./src/util');
 const { ensureThumbnailCache } = require('./src/endpoints/thumbnails');
-const { loadTokenizers } = require('./src/endpoints/tokenizers');
 
 // Work around a node v20.0.0, v20.1.0, and v20.2.0 bug. The issue was fixed in v20.3.0.
 // https://github.com/nodejs/node/issues/47822#issuecomment-1564708870
@@ -543,22 +542,12 @@ const setupTasks = async function () {
     }
     console.log();
 
-    // TODO: do endpoint init functions depend on certain directories existing or not existing? They should be callable
-    // in any order for encapsulation reasons, but right now it's unknown if that would break anything.
-    await userModule.initUserStorage(dataRoot);
-
-    if (listen && !basicAuthMode && enableAccounts) {
-        await userModule.checkAccountsProtection();
-    }
-
-    await settingsEndpoint.init();
-    const directories = await userModule.ensurePublicDirectoriesExist();
-    await userModule.migrateUserData();
+    const directories = await userModule.getUserDirectoriesList();
     await contentManager.checkForNewContent(directories);
     await ensureThumbnailCache();
     cleanUploads();
 
-    await loadTokenizers();
+    await settingsEndpoint.init();
     await statsEndpoint.init();
 
     const cleanupPlugins = await loadPlugins();
@@ -581,7 +570,6 @@ const setupTasks = async function () {
         exitProcess();
     });
 
-
     console.log('Launching...');
 
     if (autorun) open(autorunUrl.toString());
@@ -601,6 +589,9 @@ const setupTasks = async function () {
         }
     }
 
+    if (listen && !basicAuthMode && enableAccounts) {
+        await userModule.checkAccountsProtection();
+    }
 };
 
 /**
@@ -642,21 +633,27 @@ function setWindowTitle(title) {
     }
 }
 
-if (cliArguments.ssl) {
-    https.createServer(
-        {
-            cert: fs.readFileSync(cliArguments.certPath),
-            key: fs.readFileSync(cliArguments.keyPath),
-        }, app)
-        .listen(
-            Number(tavernUrl.port) || 443,
-            tavernUrl.hostname,
-            setupTasks,
-        );
-} else {
-    http.createServer(app).listen(
-        Number(tavernUrl.port) || 80,
-        tavernUrl.hostname,
-        setupTasks,
-    );
-}
+// User storage module needs to be initialized before starting the server
+userModule.initUserStorage(dataRoot)
+    .then(userModule.ensurePublicDirectoriesExist)
+    .then(userModule.migrateUserData)
+    .finally(() => {
+        if (cliArguments.ssl) {
+            https.createServer(
+                {
+                    cert: fs.readFileSync(cliArguments.certPath),
+                    key: fs.readFileSync(cliArguments.keyPath),
+                }, app)
+                .listen(
+                    Number(tavernUrl.port) || 443,
+                    tavernUrl.hostname,
+                    setupTasks,
+                );
+        } else {
+            http.createServer(app).listen(
+                Number(tavernUrl.port) || 80,
+                tavernUrl.hostname,
+                setupTasks,
+            );
+        }
+    });

src/endpoints/content-manager.js

@@ -386,6 +386,40 @@ async function downloadJannyCharacter(uuid) {
     throw new Error('Failed to download character');
 }
 
+//Download Character Cards from AICharactersCards.com (AICC) API.
+async function downloadAICCCharacter(id) {
+    const apiURL = `https://aicharactercards.com/wp-json/pngapi/v1/image/${id}`;
+    try {
+        const response = await fetch(apiURL);
+        if (!response.ok) {
+            throw new Error(`Failed to download character: ${response.statusText}`);
+        }
+
+        const contentType = response.headers.get('content-type') || 'image/png'; // Default to 'image/png' if header is missing
+        const buffer = await response.buffer();
+        const fileName = `${sanitize(id)}.png`; // Assuming PNG, but adjust based on actual content or headers
+
+        return {
+            buffer: buffer,
+            fileName: fileName,
+            fileType: contentType
+        };
+    } catch (error) {
+        console.error('Error downloading character:', error);
+        throw error;
+    }
+}
+
+function parseAICC(url) {
+    const pattern = /^https?:\/\/aicharactercards\.com\/character-cards\/([^\/]+)\/([^\/]+)\/?$|([^\/]+)\/([^\/]+)$/;
+    const match = url.match(pattern);
+    if (match) {
+        // Match group 1 & 2 for full URL, 3 & 4 for relative path
+        return match[1] && match[2] ? `${match[1]}/${match[2]}` : `${match[3]}/${match[4]}`;
+    }
+    return null;
+}
+
 /**
 * @param {String} url
 * @returns {String | null } UUID of the character
@@ -414,6 +448,7 @@ router.post('/importURL', jsonParser, async (request, response) => {
 
         const isJannnyContent = url.includes('janitorai');
         const isPygmalionContent = url.includes('pygmalion.chat');
+        const isAICharacterCardsContent = url.includes('aicharactercards.com');
 
         if (isPygmalionContent) {
             const uuid = getUuidFromUrl(url);
@@ -431,6 +466,13 @@ router.post('/importURL', jsonParser, async (request, response) => {
 
             type = 'character';
             result = await downloadJannyCharacter(uuid);
+        } else if (isAICharacterCardsContent) {
+            const AICCParsed = parseAICC(url);
+            if (!AICCParsed) {
+                return response.sendStatus(404);
+            }
+            type = 'character';
+            result = await downloadAICCCharacter(AICCParsed);
         } else {
             const chubParsed = parseChubUrl(url);
             type = chubParsed?.type;
@@ -469,6 +511,7 @@ router.post('/importUUID', jsonParser, async (request, response) => {
 
         const isJannny = uuid.includes('_character');
         const isPygmalion = (!isJannny && uuid.length == 36);
+        const isAICC = uuid.startsWith('AICC/');
         const uuidType = uuid.includes('lorebook') ? 'lorebook' : 'character';
 
         if (isPygmalion) {
@@ -477,6 +520,10 @@ router.post('/importUUID', jsonParser, async (request, response) => {
         } else if (isJannny) {
             console.log('Downloading Janitor character:', uuid.split('_')[0]);
             result = await downloadJannyCharacter(uuid.split('_')[0]);
+        } else if (isAICC) {
+            const [, author, card] = uuid.split('/');
+            console.log('Downloading AICC character:', `${author}/${card}`);
+            result = await downloadAICCCharacter(`${author}/${card}`);
         } else {
             if (uuidType === 'character') {
                 console.log('Downloading chub character:', uuid);

src/endpoints/tokenizers.js

@@ -10,6 +10,10 @@ const { TEXTGEN_TYPES } = require('../constants');
 const { jsonParser } = require('../express-common');
 const { setAdditionalHeaders } = require('../additional-headers');
 
+/**
+ * @typedef { (req: import('express').Request, res: import('express').Response) => Promise<any> } TokenizationHandler
+ */
+
 /**
  * @type {{[key: string]: import("@dqbd/tiktoken").Tiktoken}} Tokenizers cache
  */
@@ -48,16 +52,30 @@ const TEXT_COMPLETION_MODELS = [
 
 const CHARS_PER_TOKEN = 3.35;
 
+/**
+ * Sentencepiece tokenizer for tokenizing text.
+ */
 class SentencePieceTokenizer {
+    /**
+     * @type {import('@agnai/sentencepiece-js').SentencePieceProcessor} Sentencepiece tokenizer instance
+     */
     #instance;
+    /**
+     * @type {string} Path to the tokenizer model
+     */
     #model;
 
+    /**
+     * Creates a new Sentencepiece tokenizer.
+     * @param {string} model Path to the tokenizer model
+     */
     constructor(model) {
         this.#model = model;
     }
 
     /**
      * Gets the Sentencepiece tokenizer instance.
+     * @returns {Promise<import('@agnai/sentencepiece-js').SentencePieceProcessor|null>} Sentencepiece tokenizer instance
      */
     async get() {
         if (this.#instance) {
@@ -76,18 +94,61 @@ class SentencePieceTokenizer {
     }
 }
 
-const spp_llama = new SentencePieceTokenizer('src/sentencepiece/llama.model');
-const spp_nerd = new SentencePieceTokenizer('src/sentencepiece/nerdstash.model');
-const spp_nerd_v2 = new SentencePieceTokenizer('src/sentencepiece/nerdstash_v2.model');
-const spp_mistral = new SentencePieceTokenizer('src/sentencepiece/mistral.model');
-const spp_yi = new SentencePieceTokenizer('src/sentencepiece/yi.model');
-let claude_tokenizer;
+/**
+ * Web tokenizer for tokenizing text.
+ */
+class WebTokenizer {
+    /**
+     * @type {Tokenizer} Web tokenizer instance
+     */
+    #instance;
+    /**
+     * @type {string} Path to the tokenizer model
+     */
+    #model;
+
+    /**
+     * Creates a new Web tokenizer.
+     * @param {string} model Path to the tokenizer model
+     */
+    constructor(model) {
+        this.#model = model;
+    }
+
+    /**
+     * Gets the Web tokenizer instance.
+     * @returns {Promise<Tokenizer|null>} Web tokenizer instance
+     */
+    async get() {
+        if (this.#instance) {
+            return this.#instance;
+        }
+
+        try {
+            const arrayBuffer = fs.readFileSync(this.#model).buffer;
+            this.#instance = await Tokenizer.fromJSON(arrayBuffer);
+            console.log('Instantiated the tokenizer for', path.parse(this.#model).name);
+            return this.#instance;
+        } catch (error) {
+            console.error('Web tokenizer failed to load: ' + this.#model, error);
+            return null;
+        }
+    }
+}
+
+const spp_llama = new SentencePieceTokenizer('src/tokenizers/llama.model');
+const spp_nerd = new SentencePieceTokenizer('src/tokenizers/nerdstash.model');
+const spp_nerd_v2 = new SentencePieceTokenizer('src/tokenizers/nerdstash_v2.model');
+const spp_mistral = new SentencePieceTokenizer('src/tokenizers/mistral.model');
+const spp_yi = new SentencePieceTokenizer('src/tokenizers/yi.model');
+const claude_tokenizer = new WebTokenizer('src/tokenizers/claude.json');
 
 const sentencepieceTokenizers = [
     'llama',
     'nerdstash',
     'nerdstash_v2',
     'mistral',
+    'yi',
 ];
 
 /**
@@ -112,6 +173,10 @@ function getSentencepiceTokenizer(model) {
         return spp_nerd_v2;
     }
 
+    if (model.includes('yi')) {
+        return spp_yi;
+    }
+
     return null;
 }
 
@@ -168,13 +233,23 @@ async function getTiktokenChunks(tokenizer, ids) {
     return chunks;
 }
 
-async function getWebTokenizersChunks(tokenizer, ids) {
+/**
+ * Gets the token chunks for the given token IDs using the Web tokenizer.
+ * @param {Tokenizer} tokenizer Web tokenizer instance
+ * @param {number[]} ids Token IDs
+ * @returns {string[]} Token chunks
+ */
+function getWebTokenizersChunks(tokenizer, ids) {
     const chunks = [];
 
-    for (let i = 0; i < ids.length; i++) {
-        const id = ids[i];
-        const chunkText = await tokenizer.decode(new Uint32Array([id]));
+    for (let i = 0, lastProcessed = 0; i < ids.length; i++) {
+        const chunkIds = ids.slice(lastProcessed, i + 1);
+        const chunkText = tokenizer.decode(new Int32Array(chunkIds));
+        if (chunkText === '<27>') {
+            continue;
+        }
         chunks.push(chunkText);
+        lastProcessed = i + 1;
     }
 
     return chunks;
@@ -237,17 +312,12 @@ function getTiktokenTokenizer(model) {
     return tokenizer;
 }
 
-async function loadClaudeTokenizer(modelPath) {
-    try {
-        const arrayBuffer = fs.readFileSync(modelPath).buffer;
-        const instance = await Tokenizer.fromJSON(arrayBuffer);
-        return instance;
-    } catch (error) {
-        console.error('Claude tokenizer failed to load: ' + modelPath, error);
-        return null;
-    }
-}
-
+/**
+ * Counts the tokens for the given messages using the Claude tokenizer.
+ * @param {Tokenizer} tokenizer Web tokenizer
+ * @param {object[]} messages Array of messages
+ * @returns {number} Number of tokens
+ */
 function countClaudeTokens(tokenizer, messages) {
     // Should be fine if we use the old conversion method instead of the messages API one i think?
     const convertedPrompt = convertClaudePrompt(messages, false, '', false, false, '', false);
@@ -264,9 +334,14 @@ function countClaudeTokens(tokenizer, messages) {
 /**
  * Creates an API handler for encoding Sentencepiece tokens.
  * @param {SentencePieceTokenizer} tokenizer Sentencepiece tokenizer
- * @returns {any} Handler function
+ * @returns {TokenizationHandler} Handler function
  */
 function createSentencepieceEncodingHandler(tokenizer) {
+    /**
+     * Request handler for encoding Sentencepiece tokens.
+     * @param {import('express').Request} request
+     * @param {import('express').Response} response
+     */
     return async function (request, response) {
         try {
             if (!request.body) {
@@ -276,7 +351,7 @@ function createSentencepieceEncodingHandler(tokenizer) {
             const text = request.body.text || '';
             const instance = await tokenizer?.get();
             const { ids, count } = await countSentencepieceTokens(tokenizer, text);
-            const chunks = await instance?.encodePieces(text);
+            const chunks = instance?.encodePieces(text);
             return response.send({ ids, count, chunks });
         } catch (error) {
             console.log(error);
@@ -288,9 +363,14 @@ function createSentencepieceEncodingHandler(tokenizer) {
 /**
  * Creates an API handler for decoding Sentencepiece tokens.
  * @param {SentencePieceTokenizer} tokenizer Sentencepiece tokenizer
- * @returns {any} Handler function
+ * @returns {TokenizationHandler} Handler function
  */
 function createSentencepieceDecodingHandler(tokenizer) {
+    /**
+     * Request handler for decoding Sentencepiece tokens.
+     * @param {import('express').Request} request
+     * @param {import('express').Response} response
+     */
     return async function (request, response) {
         try {
             if (!request.body) {
@@ -299,6 +379,7 @@ function createSentencepieceDecodingHandler(tokenizer) {
 
             const ids = request.body.ids || [];
             const instance = await tokenizer?.get();
+            if (!instance) throw new Error('Failed to load the Sentencepiece tokenizer');
             const ops = ids.map(id => instance.decodeIds([id]));
             const chunks = await Promise.all(ops);
             const text = chunks.join('');
@@ -313,9 +394,14 @@ function createSentencepieceDecodingHandler(tokenizer) {
 /**
  * Creates an API handler for encoding Tiktoken tokens.
  * @param {string} modelId Tiktoken model ID
- * @returns {any} Handler function
+ * @returns {TokenizationHandler} Handler function
  */
 function createTiktokenEncodingHandler(modelId) {
+    /**
+     * Request handler for encoding Tiktoken tokens.
+     * @param {import('express').Request} request
+     * @param {import('express').Response} response
+     */
     return async function (request, response) {
         try {
             if (!request.body) {
@@ -337,9 +423,14 @@ function createTiktokenEncodingHandler(modelId) {
 /**
  * Creates an API handler for decoding Tiktoken tokens.
  * @param {string} modelId Tiktoken model ID
- * @returns {any} Handler function
+ * @returns {TokenizationHandler} Handler function
  */
 function createTiktokenDecodingHandler(modelId) {
+    /**
+     * Request handler for decoding Tiktoken tokens.
+     * @param {import('express').Request} request
+     * @param {import('express').Response} response
+     */
     return async function (request, response) {
         try {
             if (!request.body) {
@@ -358,14 +449,6 @@ function createTiktokenDecodingHandler(modelId) {
     };
 }
 
-/**
- * Loads the model tokenizers.
- * @returns {Promise<void>} Promise that resolves when the tokenizers are loaded
- */
-async function loadTokenizers() {
-    claude_tokenizer = await loadClaudeTokenizer('src/claude.json');
-}
-
 const router = express.Router();
 
 router.post('/ai21/count', jsonParser, async function (req, res) {
@@ -446,8 +529,10 @@ router.post('/openai/encode', jsonParser, async function (req, res) {
 
         if (queryModel.includes('claude')) {
             const text = req.body.text || '';
-            const tokens = Object.values(claude_tokenizer.encode(text));
-            const chunks = await getWebTokenizersChunks(claude_tokenizer, tokens);
+            const instance = await claude_tokenizer.get();
+            if (!instance) throw new Error('Failed to load the Claude tokenizer');
+            const tokens = Object.values(instance.encode(text));
+            const chunks = getWebTokenizersChunks(instance, tokens);
             return res.send({ ids: tokens, count: tokens.length, chunks });
         }
 
@@ -481,7 +566,9 @@ router.post('/openai/decode', jsonParser, async function (req, res) {
 
         if (queryModel.includes('claude')) {
             const ids = req.body.ids || [];
-            const chunkText = await claude_tokenizer.decode(new Uint32Array(ids));
+            const instance = await claude_tokenizer.get();
+            if (!instance) throw new Error('Failed to load the Claude tokenizer');
+            const chunkText = instance.decode(new Int32Array(ids));
             return res.send({ text: chunkText });
         }
 
@@ -503,7 +590,9 @@ router.post('/openai/count', jsonParser, async function (req, res) {
         const model = getTokenizerModel(queryModel);
 
         if (model === 'claude') {
-            num_tokens = countClaudeTokens(claude_tokenizer, req.body);
+            const instance = await claude_tokenizer.get();
+            if (!instance) throw new Error('Failed to load the Claude tokenizer');
+            num_tokens = countClaudeTokens(instance, req.body);
             return res.send({ 'token_count': num_tokens });
         }
 
@@ -665,7 +754,6 @@ module.exports = {
     getTokenizerModel,
     getTiktokenTokenizer,
     countClaudeTokens,
-    loadTokenizers,
     getSentencepiceTokenizer,
     sentencepieceTokenizers,
     router,

src/users.js

@@ -112,6 +112,16 @@ async function ensurePublicDirectoriesExist() {
     return directoriesList;
 }
 
+/**
+ * Gets a list of all user directories.
+ * @returns {Promise<import('./users').UserDirectoryList[]>} - The list of user directories
+ */
+async function getUserDirectoriesList() {
+    const userHandles = await getAllUserHandles();
+    const directoriesList = userHandles.map(handle => getUserDirectories(handle));
+    return directoriesList;
+}
+
 /**
  * Perform migration from the old user data format to the new one.
  */
@@ -707,6 +717,7 @@ module.exports = {
     toAvatarKey,
     initUserStorage,
     ensurePublicDirectoriesExist,
+    getUserDirectoriesList,
     getAllUserHandles,
     getUserDirectories,
     setUserDataMiddleware,