💫 feat: Config File & Custom Endpoints

.gitignore

@@ -48,6 +48,9 @@ bower_components/
 .floo
 .flooignore
 
+#config file
+librechat.yaml
+
 # Environment
 .npmrc
 .env*

api/app/clients/BaseClient.js

@@ -520,6 +520,7 @@ class BaseClient {
     await saveConvo(user, {
       conversationId: message.conversationId,
       endpoint: this.options.endpoint,
+      endpointType: this.options.endpointType,
       ...endpointOptions,
     });
   }

api/app/clients/OpenAIClient.js

@@ -1,6 +1,6 @@
 const OpenAI = require('openai');
 const { HttpsProxyAgent } = require('https-proxy-agent');
-const { getResponseSender, EModelEndpoint } = require('librechat-data-provider');
+const { getResponseSender } = require('librechat-data-provider');
 const { encoding_for_model: encodingForModel, get_encoding: getEncoding } = require('tiktoken');
 const { encodeAndFormat, validateVisionModel } = require('~/server/services/Files/images');
 const { getModelMaxTokens, genAzureChatCompletion, extractBaseURL } = require('~/utils');
@@ -94,10 +94,23 @@ class OpenAIClient extends BaseClient {
     }
 
     const { reverseProxyUrl: reverseProxy } = this.options;
+
+    if (
+      !this.useOpenRouter &&
+      reverseProxy &&
+      reverseProxy.includes('https://openrouter.ai/api/v1')
+    ) {
+      this.useOpenRouter = true;
+    }
+
     this.FORCE_PROMPT =
       isEnabled(OPENAI_FORCE_PROMPT) ||
       (reverseProxy && reverseProxy.includes('completions') && !reverseProxy.includes('chat'));
 
+    if (typeof this.options.forcePrompt === 'boolean') {
+      this.FORCE_PROMPT = this.options.forcePrompt;
+    }
+
     if (this.azure && process.env.AZURE_OPENAI_DEFAULT_MODEL) {
       this.azureEndpoint = genAzureChatCompletion(this.azure, this.modelOptions.model);
       this.modelOptions.model = process.env.AZURE_OPENAI_DEFAULT_MODEL;
@@ -146,8 +159,10 @@ class OpenAIClient extends BaseClient {
       this.options.sender ??
       getResponseSender({
         model: this.modelOptions.model,
-        endpoint: EModelEndpoint.openAI,
+        endpoint: this.options.endpoint,
+        endpointType: this.options.endpointType,
         chatGptLabel: this.options.chatGptLabel,
+        modelDisplayLabel: this.options.modelDisplayLabel,
       });
 
     this.userLabel = this.options.userLabel || 'User';
@@ -434,7 +449,7 @@ class OpenAIClient extends BaseClient {
         },
         opts.abortController || new AbortController(),
       );
-    } else if (typeof opts.onProgress === 'function') {
+    } else if (typeof opts.onProgress === 'function' || this.options.useChatCompletion) {
       reply = await this.chatCompletion({
         payload,
         clientOptions: opts,
@@ -530,6 +545,19 @@ class OpenAIClient extends BaseClient {
     return llm;
   }
 
+  /**
+   * Generates a concise title for a conversation based on the user's input text and response.
+   * Uses either specified method or starts with the OpenAI `functions` method (using LangChain).
+   * If the `functions` method fails, it falls back to the `completion` method,
+   * which involves sending a chat completion request with specific instructions for title generation.
+   *
+   * @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 = `||>User:
@@ -539,32 +567,25 @@ class OpenAIClient extends BaseClient {
 
     const { OPENAI_TITLE_MODEL } = process.env ?? {};
 
+    const model = this.options.titleModel ?? OPENAI_TITLE_MODEL ?? 'gpt-3.5-turbo';
+
     const modelOptions = {
-      model: OPENAI_TITLE_MODEL ?? 'gpt-3.5-turbo',
+      // TODO: remove the gpt fallback and make it specific to endpoint
+      model,
       temperature: 0.2,
       presence_penalty: 0,
       frequency_penalty: 0,
       max_tokens: 16,
     };
 
-    try {
-      this.abortController = new AbortController();
-      const llm = this.initializeLLM({ ...modelOptions, context: 'title', tokenBuffer: 150 });
-      title = await runTitleChain({ llm, text, convo, signal: this.abortController.signal });
-    } catch (e) {
-      if (e?.message?.toLowerCase()?.includes('abort')) {
-        logger.debug('[OpenAIClient] Aborted title generation');
-        return;
-      }
-      logger.error(
-        '[OpenAIClient] There was an issue generating title with LangChain, trying the old method...',
-        e,
-      );
-      modelOptions.model = OPENAI_TITLE_MODEL ?? 'gpt-3.5-turbo';
+    const titleChatCompletion = async () => {
+      modelOptions.model = model;
+
       if (this.azure) {
         modelOptions.model = process.env.AZURE_OPENAI_DEFAULT_MODEL ?? modelOptions.model;
         this.azureEndpoint = genAzureChatCompletion(this.azure, modelOptions.model);
       }
+
       const instructionsPayload = [
         {
           role: 'system',
@@ -578,10 +599,38 @@ ${convo}
       ];
 
       try {
-        title = (await this.sendPayload(instructionsPayload, { modelOptions })).replaceAll('"', '');
+        title = (
+          await this.sendPayload(instructionsPayload, { modelOptions, useChatCompletion: true })
+        ).replaceAll('"', '');
       } catch (e) {
-        logger.error('[OpenAIClient] There was another issue generating the title', e);
+        logger.error(
+          '[OpenAIClient] There was an issue generating the title with the completion method',
+          e,
+        );
+      }
+    };
+
+    if (this.options.titleMethod === 'completion') {
+      await titleChatCompletion();
+      logger.debug('[OpenAIClient] Convo Title: ' + title);
+      return title;
+    }
+
+    try {
+      this.abortController = new AbortController();
+      const llm = this.initializeLLM({ ...modelOptions, context: 'title', tokenBuffer: 150 });
+      title = await runTitleChain({ llm, text, convo, signal: this.abortController.signal });
+    } catch (e) {
+      if (e?.message?.toLowerCase()?.includes('abort')) {
+        logger.debug('[OpenAIClient] Aborted title generation');
+        return;
       }
+      logger.error(
+        '[OpenAIClient] There was an issue generating title with LangChain, trying completion method...',
+        e,
+      );
+
+      await titleChatCompletion();
     }
 
     logger.debug('[OpenAIClient] Convo Title: ' + title);
@@ -593,8 +642,11 @@ ${convo}
     let context = messagesToRefine;
     let prompt;
 
+    // TODO: remove the gpt fallback and make it specific to endpoint
     const { OPENAI_SUMMARY_MODEL = 'gpt-3.5-turbo' } = process.env ?? {};
-    const maxContextTokens = getModelMaxTokens(OPENAI_SUMMARY_MODEL) ?? 4095;
+    const model = this.options.summaryModel ?? OPENAI_SUMMARY_MODEL;
+    const maxContextTokens = getModelMaxTokens(model) ?? 4095;
+
     // 3 tokens for the assistant label, and 98 for the summarizer prompt (101)
     let promptBuffer = 101;
 
@@ -644,7 +696,7 @@ ${convo}
     logger.debug('[OpenAIClient] initialPromptTokens', initialPromptTokens);
 
     const llm = this.initializeLLM({
-      model: OPENAI_SUMMARY_MODEL,
+      model,
       temperature: 0.2,
       context: 'summary',
       tokenBuffer: initialPromptTokens,
@@ -719,7 +771,9 @@ ${convo}
       if (!abortController) {
         abortController = new AbortController();
       }
-      const modelOptions = { ...this.modelOptions };
+
+      let modelOptions = { ...this.modelOptions };
+
       if (typeof onProgress === 'function') {
         modelOptions.stream = true;
       }
@@ -779,6 +833,27 @@ ${convo}
         ...opts,
       });
 
+      /* hacky fix for Mistral AI API not allowing a singular system message in payload */
+      if (opts.baseURL.includes('https://api.mistral.ai/v1') && modelOptions.messages) {
+        const { messages } = modelOptions;
+        if (messages.length === 1 && messages[0].role === 'system') {
+          modelOptions.messages[0].role = 'user';
+        }
+      }
+
+      if (this.options.addParams && typeof this.options.addParams === 'object') {
+        modelOptions = {
+          ...modelOptions,
+          ...this.options.addParams,
+        };
+      }
+
+      if (this.options.dropParams && Array.isArray(this.options.dropParams)) {
+        this.options.dropParams.forEach((param) => {
+          delete modelOptions[param];
+        });
+      }
+
       let UnexpectedRoleError = false;
       if (modelOptions.stream) {
         const stream = await openai.beta.chat.completions
@@ -859,7 +934,6 @@ ${convo}
         (err instanceof OpenAI.OpenAIError && err?.message?.includes('missing finish_reason'))
       ) {
         logger.error('[OpenAIClient] Known OpenAI error:', err);
-        await abortController.abortCompletion();
         return intermediateReply;
       } else if (err instanceof OpenAI.APIError) {
         if (intermediateReply) {

api/cache/getCustomConfig.js

@@ -0,0 +1,23 @@
+const { CacheKeys } = require('librechat-data-provider');
+const loadCustomConfig = require('~/server/services/Config/loadCustomConfig');
+const getLogStores = require('./getLogStores');
+
+/**
+ * Retrieves the configuration object
+ * @function getCustomConfig */
+async function getCustomConfig() {
+  const cache = getLogStores(CacheKeys.CONFIG_STORE);
+  let customConfig = await cache.get(CacheKeys.CUSTOM_CONFIG);
+
+  if (!customConfig) {
+    customConfig = await loadCustomConfig();
+  }
+
+  if (!customConfig) {
+    return null;
+  }
+
+  return customConfig;
+}
+
+module.exports = getCustomConfig;

api/cache/getLogStores.js

@@ -1,9 +1,10 @@
 const Keyv = require('keyv');
-const keyvMongo = require('./keyvMongo');
-const keyvRedis = require('./keyvRedis');
-const { CacheKeys } = require('~/common/enums');
-const { math, isEnabled } = require('~/server/utils');
+const { CacheKeys } = require('librechat-data-provider');
 const { logFile, violationFile } = require('./keyvFiles');
+const { math, isEnabled } = require('~/server/utils');
+const keyvRedis = require('./keyvRedis');
+const keyvMongo = require('./keyvMongo');
+
 const { BAN_DURATION, USE_REDIS } = process.env ?? {};
 
 const duration = math(BAN_DURATION, 7200000);
@@ -20,10 +21,10 @@ const pending_req = isEnabled(USE_REDIS)
 
 const config = isEnabled(USE_REDIS)
   ? new Keyv({ store: keyvRedis })
-  : new Keyv({ namespace: CacheKeys.CONFIG });
+  : new Keyv({ namespace: CacheKeys.CONFIG_STORE });
 
 const namespaces = {
-  config,
+  [CacheKeys.CONFIG_STORE]: config,
   pending_req,
   ban: new Keyv({ store: keyvMongo, namespace: 'bans', ttl: duration }),
   general: new Keyv({ store: logFile, namespace: 'violations' }),
@@ -39,19 +40,15 @@ const namespaces = {
  * Returns the keyv cache specified by type.
  * If an invalid type is passed, an error will be thrown.
  *
- * @module getLogStores
- * @requires keyv - a simple key-value storage that allows you to easily switch out storage adapters.
- * @requires keyvFiles - a module that includes the logFile and violationFile.
- *
- * @param {string} type - The type of violation, which can be 'concurrent', 'message_limit', 'registrations' or 'logins'.
- * @returns {Keyv} - If a valid type is passed, returns an object containing the logs for violations of the specified type.
- * @throws Will throw an error if an invalid violation type is passed.
+ * @param {string} key - The key for the namespace to access
+ * @returns {Keyv} - If a valid key is passed, returns an object containing the cache store of the specified key.
+ * @throws Will throw an error if an invalid key is passed.
  */
-const getLogStores = (type) => {
-  if (!type || !namespaces[type]) {
-    throw new Error(`Invalid store type: ${type}`);
+const getLogStores = (key) => {
+  if (!key || !namespaces[key]) {
+    throw new Error(`Invalid store key: ${key}`);
   }
-  return namespaces[type];
+  return namespaces[key];
 };
 
 module.exports = getLogStores;

api/models/schema/convoSchema.js

@@ -18,36 +18,29 @@ const convoSchema = mongoose.Schema(
     user: {
       type: String,
       index: true,
-      // default: null,
     },
     messages: [{ type: mongoose.Schema.Types.ObjectId, ref: 'Message' }],
     // google only
-    examples: [{ type: mongoose.Schema.Types.Mixed }],
+    examples: { type: [{ type: mongoose.Schema.Types.Mixed }], default: undefined },
     agentOptions: {
       type: mongoose.Schema.Types.Mixed,
-      // default: null,
     },
     ...conversationPreset,
     // for bingAI only
     bingConversationId: {
       type: String,
-      // default: null,
     },
     jailbreakConversationId: {
       type: String,
-      // default: null,
     },
     conversationSignature: {
       type: String,
-      // default: null,
     },
     clientId: {
       type: String,
-      // default: null,
     },
     invocationId: {
       type: Number,
-      // default: 1,
     },
   },
   { timestamps: true },

api/models/schema/defaults.js

@@ -5,6 +5,9 @@ const conversationPreset = {
     default: null,
     required: true,
   },
+  endpointType: {
+    type: String,
+  },
   // for azureOpenAI, openAI, chatGPTBrowser only
   model: {
     type: String,
@@ -95,7 +98,6 @@ const agentOptions = {
     // default: null,
     required: false,
   },
-  // for google only
   modelLabel: {
     type: String,
     // default: null,