🎉 feat: Optimizations and Anthropic Title Generation (#2184)

* feat: add claude-3-haiku-20240307 to default anthropic list

* refactor: optimize `saveMessage` calls mid-stream via throttling

* chore: remove addMetadata operations and consolidate in BaseClient

* fix(listAssistantsForAzure): attempt to specify correct model mapping as accurately as possible (#2177)

* refactor(client): update last conversation setup with current assistant model, call newConvo again when assistants load to allow fast initial load and ensure assistant model is always the default, not the last selected model

* refactor(cache): explicitly add TTL of 2 minutes when setting titleCache and add default TTL of 10 minutes to abortKeys cache

* feat(AnthropicClient): conversation titling using Anthropic Function Calling

* chore: remove extraneous token usage logging

* fix(convos): unhandled edge case for conversation grouping (undefined conversation)

* style: Improved style of Search Bar after recent UI update

* chore: remove unused code, content part helpers

* feat: always show code option

api/app/clients/AnthropicClient.js

@@ -6,7 +6,13 @@ const {
   validateVisionModel,
 } = require('librechat-data-provider');
 const { encodeAndFormat } = require('~/server/services/Files/images/encode');
-const { formatMessage, createContextHandlers } = require('./prompts');
+const {
+  titleFunctionPrompt,
+  parseTitleFromPrompt,
+  truncateText,
+  formatMessage,
+  createContextHandlers,
+} = require('./prompts');
 const spendTokens = require('~/models/spendTokens');
 const { getModelMaxTokens } = require('~/utils');
 const BaseClient = require('./BaseClient');
@@ -108,7 +114,12 @@ class AnthropicClient extends BaseClient {
     return this;
   }
 
+  /**
+   * Get the initialized Anthropic client.
+   * @returns {Anthropic} The Anthropic client instance.
+   */
   getClient() {
+    /** @type {Anthropic.default.RequestOptions} */
     const options = {
       apiKey: this.apiKey,
     };
@@ -176,14 +187,13 @@ class AnthropicClient extends BaseClient {
     return files;
   }
 
-  async recordTokenUsage({ promptTokens, completionTokens }) {
-    logger.debug('[AnthropicClient] recordTokenUsage:', { promptTokens, completionTokens });
+  async recordTokenUsage({ promptTokens, completionTokens, model, context = 'message' }) {
     await spendTokens(
       {
+        context,
         user: this.user,
-        model: this.modelOptions.model,
-        context: 'message',
         conversationId: this.conversationId,
+        model: model ?? this.modelOptions.model,
         endpointTokenConfig: this.options.endpointTokenConfig,
       },
       { promptTokens, completionTokens },
@@ -512,8 +522,15 @@ class AnthropicClient extends BaseClient {
     logger.debug('AnthropicClient doesn\'t use getCompletion (all handled in sendCompletion)');
   }
 
-  async createResponse(client, options) {
-    return this.useMessages
+  /**
+   * Creates a message or completion response using the Anthropic client.
+   * @param {Anthropic} client - The Anthropic client instance.
+   * @param {Anthropic.default.MessageCreateParams | Anthropic.default.CompletionCreateParams} options - The options for the message or completion.
+   * @param {boolean} useMessages - Whether to use messages or completions. Defaults to `this.useMessages`.
+   * @returns {Promise<Anthropic.default.Message | Anthropic.default.Completion>} The response from the Anthropic client.
+   */
+  async createResponse(client, options, useMessages) {
+    return useMessages ?? this.useMessages
       ? await client.messages.create(options)
       : await client.completions.create(options);
   }
@@ -663,6 +680,78 @@ class AnthropicClient extends BaseClient {
   getTokenCount(text) {
     return this.gptEncoder.encode(text, 'all').length;
   }
+
+  /**
+   * Generates a concise title for a conversation based on the user's input text and response.
+   * Involves sending a chat completion request with specific instructions for title generation.
+   *
+   * This function capitlizes on [Anthropic's function calling training](https://docs.anthropic.com/claude/docs/functions-external-tools).
+   *
+   * @param {Object} params - The parameters for the conversation title generation.
+   * @param {string} params.text - The user's input.
+   * @param {string} [params.responseText=''] - The AI's immediate response to the user.
+   *
+   * @returns {Promise<string | 'New Chat'>} A promise that resolves to the generated conversation title.
+   *                            In case of failure, it will return the default title, "New Chat".
+   */
+  async titleConvo({ text, responseText = '' }) {
+    let title = 'New Chat';
+    const convo = `<initial_message>
+  ${truncateText(text)}
+  </initial_message>
+  <response>
+  ${JSON.stringify(truncateText(responseText))}
+  </response>`;
+
+    const { ANTHROPIC_TITLE_MODEL } = process.env ?? {};
+    const model = this.options.titleModel ?? ANTHROPIC_TITLE_MODEL ?? 'claude-3-haiku-20240307';
+    const system = titleFunctionPrompt;
+
+    const titleChatCompletion = async () => {
+      const content = `<conversation_context>
+  ${convo}
+  </conversation_context>
+  
+  Please generate a title for this conversation.`;
+
+      const titleMessage = { role: 'user', content };
+      const requestOptions = {
+        model,
+        temperature: 0.3,
+        max_tokens: 1024,
+        system,
+        stop_sequences: ['\n\nHuman:', '\n\nAssistant', '</function_calls>'],
+        messages: [titleMessage],
+      };
+
+      try {
+        const response = await this.createResponse(this.getClient(), requestOptions, true);
+        let promptTokens = response?.usage?.input_tokens;
+        let completionTokens = response?.usage?.output_tokens;
+        if (!promptTokens) {
+          promptTokens = this.getTokenCountForMessage(titleMessage);
+          promptTokens += this.getTokenCountForMessage({ role: 'system', content: system });
+        }
+        if (!completionTokens) {
+          completionTokens = this.getTokenCountForMessage(response.content[0]);
+        }
+        await this.recordTokenUsage({
+          model,
+          promptTokens,
+          completionTokens,
+          context: 'title',
+        });
+        const text = response.content[0].text;
+        title = parseTitleFromPrompt(text);
+      } catch (e) {
+        logger.error('[AnthropicClient] There was an issue generating the title', e);
+      }
+    };
+
+    await titleChatCompletion();
+    logger.debug('[AnthropicClient] Convo Title: ' + title);
+    return title;
+  }
 }
 
 module.exports = AnthropicClient;

api/app/clients/BaseClient.js

@@ -456,6 +456,7 @@ class BaseClient {
       sender: this.sender,
       text: addSpaceIfNeeded(generation) + completion,
       promptTokens,
+      ...(this.metadata ?? {}),
     };
 
     if (

api/app/clients/OpenAIClient.js

@@ -47,6 +47,7 @@ class OpenAIClient extends BaseClient {
     /** @type {AzureOptions} */
     this.azure = options.azure || false;
     this.setOptions(options);
+    this.metadata = {};
   }
 
   // TODO: PluginsClient calls this 3x, unneeded
@@ -574,7 +575,6 @@ class OpenAIClient extends BaseClient {
     } else if (typeof opts.onProgress === 'function' || this.options.useChatCompletion) {
       reply = await this.chatCompletion({
         payload,
-        clientOptions: opts,
         onProgress: opts.onProgress,
         abortController: opts.abortController,
       });
@@ -594,9 +594,9 @@ class OpenAIClient extends BaseClient {
       }
     }
 
-    if (streamResult && typeof opts.addMetadata === 'function') {
+    if (streamResult) {
       const { finish_reason } = streamResult.choices[0];
-      opts.addMetadata({ finish_reason });
+      this.metadata = { finish_reason };
     }
     return (reply ?? '').trim();
   }
@@ -921,7 +921,6 @@ ${convo}
   }
 
   async recordTokenUsage({ promptTokens, completionTokens }) {
-    logger.debug('[OpenAIClient] recordTokenUsage:', { promptTokens, completionTokens });
     await spendTokens(
       {
         user: this.user,
@@ -941,7 +940,7 @@ ${convo}
     });
   }
 
-  async chatCompletion({ payload, onProgress, clientOptions, abortController = null }) {
+  async chatCompletion({ payload, onProgress, abortController = null }) {
     let error = null;
     const errorCallback = (err) => (error = err);
     let intermediateReply = '';
@@ -962,15 +961,6 @@ ${convo}
       }
 
       const baseURL = extractBaseURL(this.completionsUrl);
-      // let { messages: _msgsToLog, ...modelOptionsToLog } = modelOptions;
-      // if (modelOptionsToLog.messages) {
-      //   _msgsToLog = modelOptionsToLog.messages.map((msg) => {
-      //     let { content, ...rest } = msg;
-
-      //     if (content)
-      //     return { ...rest, content: truncateText(content) };
-      //   });
-      // }
       logger.debug('[OpenAIClient] chatCompletion', { baseURL, modelOptions });
       const opts = {
         baseURL,
@@ -1163,8 +1153,8 @@ ${convo}
       }
 
       const { message, finish_reason } = chatCompletion.choices[0];
-      if (chatCompletion && typeof clientOptions.addMetadata === 'function') {
-        clientOptions.addMetadata({ finish_reason });
+      if (chatCompletion) {
+        this.metadata = { finish_reason };
       }
 
       logger.debug('[OpenAIClient] chatCompletion response', chatCompletion);

api/app/clients/prompts/titlePrompts.js

@@ -27,7 +27,60 @@ ${convo}`,
   return titlePrompt;
 };
 
+const titleFunctionPrompt = `In this environment you have access to a set of tools you can use to generate the conversation title.
+  
+You may call them like this:
+<function_calls>
+<invoke>
+<tool_name>$TOOL_NAME</tool_name>
+<parameters>
+<$PARAMETER_NAME>$PARAMETER_VALUE</$PARAMETER_NAME>
+...
+</parameters>
+</invoke>
+</function_calls>
+
+Here are the tools available:
+<tools>
+<tool_description>
+<tool_name>submit_title</tool_name>
+<description>
+Submit a brief title in the conversation's language, following the parameter description closely.
+</description>
+<parameters>
+<parameter>
+<name>title</name>
+<type>string</type>
+<description>A concise, 5-word-or-less title for the conversation, using its same language, with no punctuation. Apply title case conventions appropriate for the language. For English, use AP Stylebook Title Case. Never directly mention the language name or the word "title"</description>
+</parameter>
+</parameters>
+</tool_description>
+</tools>`;
+
+/**
+ * Parses titles from title functions based on the provided prompt.
+ * @param {string} prompt - The prompt containing the title function.
+ * @returns {string} The parsed title. "New Chat" if no title is found.
+ */
+function parseTitleFromPrompt(prompt) {
+  const titleRegex = /<title>(.+?)<\/title>/;
+  const titleMatch = prompt.match(titleRegex);
+
+  if (titleMatch && titleMatch[1]) {
+    const title = titleMatch[1].trim();
+
+    // // Capitalize the first letter of each word; Note: unnecessary due to title case prompting
+    // const capitalizedTitle = title.replace(/\b\w/g, (char) => char.toUpperCase());
+
+    return title;
+  }
+
+  return 'New Chat';
+}
+
 module.exports = {
   langPrompt,
   createTitlePrompt,
+  titleFunctionPrompt,
+  parseTitleFromPrompt,
 };

api/cache/getLogStores.js

@@ -37,7 +37,7 @@ const modelQueries = isEnabled(process.env.USE_REDIS)
 
 const abortKeys = isEnabled(USE_REDIS)
   ? new Keyv({ store: keyvRedis })
-  : new Keyv({ namespace: CacheKeys.ABORT_KEYS });
+  : new Keyv({ namespace: CacheKeys.ABORT_KEYS, ttl: 600000 });
 
 const namespaces = {
   [CacheKeys.CONFIG_STORE]: config,

api/models/spendTokens.js

@@ -21,10 +21,15 @@ const { logger } = require('~/config');
  */
 const spendTokens = async (txData, tokenUsage) => {
   const { promptTokens, completionTokens } = tokenUsage;
-  logger.debug(`[spendTokens] conversationId: ${txData.conversationId} | Token usage: `, {
-    promptTokens,
-    completionTokens,
-  });
+  logger.debug(
+    `[spendTokens] conversationId: ${txData.conversationId}${
+      txData?.context ? ` | Context: ${txData?.context}` : ''
+    } | Token usage: `,
+    {
+      promptTokens,
+      completionTokens,
+    },
+  );
   let prompt, completion;
   try {
     if (promptTokens >= 0) {

api/server/controllers/AskController.js

@@ -1,3 +1,4 @@
+const throttle = require('lodash/throttle');
 const { getResponseSender, Constants } = require('librechat-data-provider');
 const { createAbortController, handleAbortError } = require('~/server/middleware');
 const { sendMessage, createOnProgress } = require('~/server/utils');
@@ -16,13 +17,10 @@ const AskController = async (req, res, next, initializeClient, addTitle) => {
 
   logger.debug('[AskController]', { text, conversationId, ...endpointOption });
 
-  let metadata;
   let userMessage;
   let promptTokens;
   let userMessageId;
   let responseMessageId;
-  let lastSavedTimestamp = 0;
-  let saveDelay = 100;
   const sender = getResponseSender({
     ...endpointOption,
     model: endpointOption.modelOptions.model,
@@ -31,8 +29,6 @@ const AskController = async (req, res, next, initializeClient, addTitle) => {
   const newConvo = !conversationId;
   const user = req.user.id;
 
-  const addMetadata = (data) => (metadata = data);
-
   const getReqData = (data = {}) => {
     for (let key in data) {
       if (key === 'userMessage') {
@@ -54,11 +50,8 @@ const AskController = async (req, res, next, initializeClient, addTitle) => {
     const { client } = await initializeClient({ req, res, endpointOption });
 
     const { onProgress: progressCallback, getPartialText } = createOnProgress({
-      onProgress: ({ text: partialText }) => {
-        const currentTimestamp = Date.now();
-
-        if (currentTimestamp - lastSavedTimestamp > saveDelay) {
-          lastSavedTimestamp = currentTimestamp;
+      onProgress: throttle(
+        ({ text: partialText }) => {
           saveMessage({
             messageId: responseMessageId,
             sender,
@@ -70,12 +63,10 @@ const AskController = async (req, res, next, initializeClient, addTitle) => {
             error: false,
             user,
           });
-        }
-
-        if (saveDelay < 500) {
-          saveDelay = 500;
-        }
-      },
+        },
+        3000,
+        { trailing: false },
+      ),
     });
 
     getText = getPartialText;
@@ -113,7 +104,6 @@ const AskController = async (req, res, next, initializeClient, addTitle) => {
       overrideParentMessageId,
       getReqData,
       onStart,
-      addMetadata,
       abortController,
       onProgress: progressCallback.call(null, {
         res,
@@ -128,10 +118,6 @@ const AskController = async (req, res, next, initializeClient, addTitle) => {
       response.parentMessageId = overrideParentMessageId;
     }
 
-    if (metadata) {
-      response = { ...response, ...metadata };
-    }
-
     response.endpoint = endpointOption.endpoint;
 
     const conversation = await getConvo(user, conversationId);

api/server/controllers/EditController.js

@@ -1,3 +1,4 @@
+const throttle = require('lodash/throttle');
 const { getResponseSender } = require('librechat-data-provider');
 const { createAbortController, handleAbortError } = require('~/server/middleware');
 const { sendMessage, createOnProgress } = require('~/server/utils');
@@ -25,11 +26,8 @@ const EditController = async (req, res, next, initializeClient) => {
     ...endpointOption,
   });
 
-  let metadata;
   let userMessage;
   let promptTokens;
-  let lastSavedTimestamp = 0;
-  let saveDelay = 100;
   const sender = getResponseSender({
     ...endpointOption,
     model: endpointOption.modelOptions.model,
@@ -38,7 +36,6 @@ const EditController = async (req, res, next, initializeClient) => {
   const userMessageId = parentMessageId;
   const user = req.user.id;
 
-  const addMetadata = (data) => (metadata = data);
   const getReqData = (data = {}) => {
     for (let key in data) {
       if (key === 'userMessage') {
@@ -53,11 +50,8 @@ const EditController = async (req, res, next, initializeClient) => {
 
   const { onProgress: progressCallback, getPartialText } = createOnProgress({
     generation,
-    onProgress: ({ text: partialText }) => {
-      const currentTimestamp = Date.now();
-
-      if (currentTimestamp - lastSavedTimestamp > saveDelay) {
-        lastSavedTimestamp = currentTimestamp;
+    onProgress: throttle(
+      ({ text: partialText }) => {
         saveMessage({
           messageId: responseMessageId,
           sender,
@@ -70,12 +64,10 @@ const EditController = async (req, res, next, initializeClient) => {
           error: false,
           user,
         });
-      }
-
-      if (saveDelay < 500) {
-        saveDelay = 500;
-      }
-    },
+      },
+      3000,
+      { trailing: false },
+    ),
   });
 
   const getAbortData = () => ({
@@ -118,7 +110,6 @@ const EditController = async (req, res, next, initializeClient) => {
       overrideParentMessageId,
       getReqData,
       onStart,
-      addMetadata,
       abortController,
       onProgress: progressCallback.call(null, {
         res,
@@ -127,10 +118,6 @@ const EditController = async (req, res, next, initializeClient) => {
       }),
     });
 
-    if (metadata) {
-      response = { ...response, ...metadata };
-    }
-
     const conversation = await getConvo(user, conversationId);
     conversation.title =
       conversation && !conversation.title ? null : conversation?.title || 'New Chat';

api/server/routes/ask/anthropic.js

@@ -1,6 +1,6 @@
 const express = require('express');
 const AskController = require('~/server/controllers/AskController');
-const { initializeClient } = require('~/server/services/Endpoints/anthropic');
+const { addTitle, initializeClient } = require('~/server/services/Endpoints/anthropic');
 const {
   setHeaders,
   handleAbort,
@@ -20,7 +20,7 @@ router.post(
   buildEndpointOption,
   setHeaders,
   async (req, res, next) => {
-    await AskController(req, res, next, initializeClient);
+    await AskController(req, res, next, initializeClient, addTitle);
   },
 );
 

api/server/routes/ask/gptPlugins.js

@@ -1,11 +1,10 @@
 const express = require('express');
-const router = express.Router();
+const throttle = require('lodash/throttle');
 const { getResponseSender, Constants } = require('librechat-data-provider');
-const { validateTools } = require('~/app');
-const { addTitle } = require('~/server/services/Endpoints/openAI');
 const { initializeClient } = require('~/server/services/Endpoints/gptPlugins');
 const { saveMessage, getConvoTitle, getConvo } = require('~/models');
 const { sendMessage, createOnProgress } = require('~/server/utils');
+const { addTitle } = require('~/server/services/Endpoints/openAI');
 const {
   handleAbort,
   createAbortController,
@@ -16,8 +15,11 @@ const {
   buildEndpointOption,
   moderateText,
 } = require('~/server/middleware');
+const { validateTools } = require('~/app');
 const { logger } = require('~/config');
 
+const router = express.Router();
+
 router.use(moderateText);
 router.post('/abort', handleAbort());
 
@@ -35,14 +37,13 @@ router.post(
       parentMessageId = null,
       overrideParentMessageId = null,
     } = req.body;
+
     logger.debug('[/ask/gptPlugins]', { text, conversationId, ...endpointOption });
-    let metadata;
+
     let userMessage;
     let promptTokens;
     let userMessageId;
     let responseMessageId;
-    let lastSavedTimestamp = 0;
-    let saveDelay = 100;
     const sender = getResponseSender({
       ...endpointOption,
       model: endpointOption.modelOptions.model,
@@ -52,7 +53,6 @@ router.post(
 
     const plugins = [];
 
-    const addMetadata = (data) => (metadata = data);
     const getReqData = (data = {}) => {
       for (let key in data) {
         if (key === 'userMessage') {
@@ -68,6 +68,7 @@ router.post(
       }
     };
 
+    const throttledSaveMessage = throttle(saveMessage, 3000, { trailing: false });
     let streaming = null;
     let timer = null;
 
@@ -77,31 +78,22 @@ router.post(
       getPartialText,
     } = createOnProgress({
       onProgress: ({ text: partialText }) => {
-        const currentTimestamp = Date.now();
-
         if (timer) {
           clearTimeout(timer);
         }
 
-        if (currentTimestamp - lastSavedTimestamp > saveDelay) {
-          lastSavedTimestamp = currentTimestamp;
-          saveMessage({
-            messageId: responseMessageId,
-            sender,
-            conversationId,
-            parentMessageId: overrideParentMessageId || userMessageId,
-            text: partialText,
-            model: endpointOption.modelOptions.model,
-            unfinished: true,
-            error: false,
-            plugins,
-            user,
-          });
-        }
-
-        if (saveDelay < 500) {
-          saveDelay = 500;
-        }
+        throttledSaveMessage({
+          messageId: responseMessageId,
+          sender,
+          conversationId,
+          parentMessageId: overrideParentMessageId || userMessageId,
+          text: partialText,
+          model: endpointOption.modelOptions.model,
+          unfinished: true,
+          error: false,
+          plugins,
+          user,
+        });
 
         streaming = new Promise((resolve) => {
           timer = setTimeout(() => {
@@ -180,7 +172,6 @@ router.post(
         onToolStart,
         onToolEnd,
         onStart,
-        addMetadata,
         getPartialText,
         ...endpointOption,
         onProgress: progressCallback.call(null, {
@@ -196,10 +187,6 @@ router.post(
         response.parentMessageId = overrideParentMessageId;
       }
 
-      if (metadata) {
-        response = { ...response, ...metadata };
-      }
-
       logger.debug('[/ask/gptPlugins]', response);
 
       response.plugins = plugins.map((p) => ({ ...p, loading: false }));

api/server/routes/edit/gptPlugins.js

@@ -1,10 +1,6 @@
 const express = require('express');
-const router = express.Router();
-const { validateTools } = require('~/app');
+const throttle = require('lodash/throttle');
 const { getResponseSender } = require('librechat-data-provider');
-const { saveMessage, getConvoTitle, getConvo } = require('~/models');
-const { initializeClient } = require('~/server/services/Endpoints/gptPlugins');
-const { sendMessage, createOnProgress, formatSteps, formatAction } = require('~/server/utils');
 const {
   handleAbort,
   createAbortController,
@@ -15,8 +11,14 @@ const {
   buildEndpointOption,
   moderateText,
 } = require('~/server/middleware');
+const { sendMessage, createOnProgress, formatSteps, formatAction } = require('~/server/utils');
+const { initializeClient } = require('~/server/services/Endpoints/gptPlugins');
+const { saveMessage, getConvoTitle, getConvo } = require('~/models');
+const { validateTools } = require('~/app');
 const { logger } = require('~/config');
 
+const router = express.Router();
+
 router.use(moderateText);
 router.post('/abort', handleAbort());
 
@@ -45,11 +47,9 @@ router.post(
       conversationId,
       ...endpointOption,
     });
-    let metadata;
+
     let userMessage;
     let promptTokens;
-    let lastSavedTimestamp = 0;
-    let saveDelay = 100;
     const sender = getResponseSender({
       ...endpointOption,
       model: endpointOption.modelOptions.model,
@@ -64,7 +64,6 @@ router.post(
       outputs: null,
     };
 
-    const addMetadata = (data) => (metadata = data);
     const getReqData = (data = {}) => {
       for (let key in data) {
         if (key === 'userMessage') {
@@ -77,6 +76,7 @@ router.post(
       }
     };
 
+    const throttledSaveMessage = throttle(saveMessage, 3000, { trailing: false });
     const {
       onProgress: progressCallback,
       sendIntermediateMessage,
@@ -84,31 +84,22 @@ router.post(
     } = createOnProgress({
       generation,
       onProgress: ({ text: partialText }) => {
-        const currentTimestamp = Date.now();
-
         if (plugin.loading === true) {
           plugin.loading = false;
         }
 
-        if (currentTimestamp - lastSavedTimestamp > saveDelay) {
-          lastSavedTimestamp = currentTimestamp;
-          saveMessage({
-            messageId: responseMessageId,
-            sender,
-            conversationId,
-            parentMessageId: overrideParentMessageId || userMessageId,
-            text: partialText,
-            model: endpointOption.modelOptions.model,
-            unfinished: true,
-            isEdited: true,
-            error: false,
-            user,
-          });
-        }
-
-        if (saveDelay < 500) {
-          saveDelay = 500;
-        }
+        throttledSaveMessage({
+          messageId: responseMessageId,
+          sender,
+          conversationId,
+          parentMessageId: overrideParentMessageId || userMessageId,
+          text: partialText,
+          model: endpointOption.modelOptions.model,
+          unfinished: true,
+          isEdited: true,
+          error: false,
+          user,
+        });
       },
     });
 
@@ -161,7 +152,6 @@ router.post(
         onAgentAction,
         onChainEnd,
         onStart,
-        addMetadata,
         ...endpointOption,
         onProgress: progressCallback.call(null, {
           res,
@@ -176,10 +166,6 @@ router.post(
         response.parentMessageId = overrideParentMessageId;
       }
 
-      if (metadata) {
-        response = { ...response, ...metadata };
-      }
-
       logger.debug('[/edit/gptPlugins] CLIENT RESPONSE', response);
       response.plugin = { ...plugin, loading: false };
       await saveMessage({ ...response, user });

api/server/services/Endpoints/anthropic/addTitle.js

@@ -0,0 +1,32 @@
+const { CacheKeys } = require('librechat-data-provider');
+const getLogStores = require('~/cache/getLogStores');
+const { isEnabled } = require('~/server/utils');
+const { saveConvo } = require('~/models');
+
+const addTitle = async (req, { text, response, client }) => {
+  const { TITLE_CONVO = 'true' } = process.env ?? {};
+  if (!isEnabled(TITLE_CONVO)) {
+    return;
+  }
+
+  if (client.options.titleConvo === false) {
+    return;
+  }
+
+  // If the request was aborted, don't generate the title.
+  if (client.abortController.signal.aborted) {
+    return;
+  }
+
+  const titleCache = getLogStores(CacheKeys.GEN_TITLE);
+  const key = `${req.user.id}-${response.conversationId}`;
+
+  const title = await client.titleConvo({ text, responseText: response?.text });
+  await titleCache.set(key, title, 120000);
+  await saveConvo(req.user.id, {
+    conversationId: response.conversationId,
+    title,
+  });
+};
+
+module.exports = addTitle;

api/server/services/Endpoints/anthropic/index.js

@@ -1,8 +1,9 @@
+const addTitle = require('./addTitle');
 const buildOptions = require('./buildOptions');
 const initializeClient = require('./initializeClient');
 
 module.exports = {
-  // addTitle, // todo
+  addTitle,
   buildOptions,
   initializeClient,
 };

api/server/services/Endpoints/assistants/addTitle.js

@@ -17,7 +17,7 @@ const addTitle = async (req, { text, responseText, conversationId, client }) =>
   const key = `${req.user.id}-${conversationId}`;
 
   const title = await client.titleConvo({ text, conversationId, responseText });
-  await titleCache.set(key, title);
+  await titleCache.set(key, title, 120000);
 
   await saveConvo(req.user.id, {
     conversationId,

api/server/services/Endpoints/assistants/index.js

@@ -35,23 +35,46 @@ const listAssistants = async ({ req, res, query }) => {
  * @returns {Promise<AssistantListResponse>} A promise that resolves to an array of assistant data merged with their respective model information.
  */
 const listAssistantsForAzure = async ({ req, res, azureConfig = {}, query }) => {
+  /** @type {Array<[string, TAzureModelConfig]>} */
+  const groupModelTuples = [];
   const promises = [];
-  const models = [];
+  /** @type {Array<TAzureGroup>} */
+  const groups = [];
 
   const { groupMap, assistantGroups } = azureConfig;
 
   for (const groupName of assistantGroups) {
     const group = groupMap[groupName];
-    req.body.model = Object.keys(group?.models)[0];
-    models.push(req.body.model);
+    groups.push(group);
+
+    const currentModelTuples = Object.entries(group?.models);
+    groupModelTuples.push(currentModelTuples);
+
+    /* The specified model is only necessary to
+    fetch assistants for the shared instance */
+    req.body.model = currentModelTuples[0][0];
     promises.push(listAssistants({ req, res, query }));
   }
 
   const resolvedQueries = await Promise.all(promises);
   const data = resolvedQueries.flatMap((res, i) =>
     res.data.map((assistant) => {
-      const model = models[i];
-      return { ...assistant, model } ?? {};
+      const deploymentName = assistant.model;
+      const currentGroup = groups[i];
+      const currentModelTuples = groupModelTuples[i];
+      const firstModel = currentModelTuples[0][0];
+
+      if (currentGroup.deploymentName === deploymentName) {
+        return { ...assistant, model: firstModel };
+      }
+
+      for (const [model, modelConfig] of currentModelTuples) {
+        if (modelConfig.deploymentName === deploymentName) {
+          return { ...assistant, model };
+        }
+      }
+
+      return { ...assistant, model: firstModel };
     }),
   );
 

api/server/services/Endpoints/openAI/addTitle.js

@@ -22,7 +22,7 @@ const addTitle = async (req, { text, response, client }) => {
   const key = `${req.user.id}-${response.conversationId}`;
 
   const title = await client.titleConvo({ text, responseText: response?.text });
-  await titleCache.set(key, title);
+  await titleCache.set(key, title, 120000);
   await saveConvo(req.user.id, {
     conversationId: response.conversationId,
     title,

api/typedefs.js

@@ -8,6 +8,12 @@
  * @memberof typedefs
  */
 
+/**
+ * @exports Anthropic
+ * @typedef {import('@anthropic-ai/sdk').default} Anthropic
+ * @memberof typedefs
+ */
+
 /**
  * @exports AssistantStreamEvent
  * @typedef {import('openai').default.Beta.AssistantStreamEvent} AssistantStreamEvent
@@ -206,6 +212,18 @@
  * @memberof typedefs
  */
 
+/**
+ * @exports TAzureModelConfig
+ * @typedef {import('librechat-data-provider').TAzureModelConfig} TAzureModelConfig
+ * @memberof typedefs
+ */
+
+/**
+ * @exports TAzureGroup
+ * @typedef {import('librechat-data-provider').TAzureGroup} TAzureGroup
+ * @memberof typedefs
+ */
+
 /**
  * @exports TAzureGroups
  * @typedef {import('librechat-data-provider').TAzureGroups} TAzureGroups