Merge branch 'main' into feat/Custom-Token-Rates-for-Endpoints

api/server/services/Config/loadConfigModels.js

@@ -47,7 +47,7 @@ async function loadConfigModels(req) {
   );
 
   /**
-   * @type {Record<string, string[]>}
+   * @type {Record<string, Promise<string[]>>}
    * Map for promises keyed by unique combination of baseURL and apiKey */
   const fetchPromisesMap = {};
   /**
@@ -102,7 +102,7 @@ async function loadConfigModels(req) {
 
     for (const name of associatedNames) {
       const endpoint = endpointsMap[name];
-      modelsConfig[name] = !modelData?.length ? endpoint.models.default ?? [] : modelData;
+      modelsConfig[name] = !modelData?.length ? (endpoint.models.default ?? []) : modelData;
     }
   }
 

api/server/services/Config/loadDefaultModels.js

@@ -5,8 +5,8 @@ const {
   getGoogleModels,
   getBedrockModels,
   getAnthropicModels,
-  getChatGPTBrowserModels,
 } = require('~/server/services/ModelService');
+const { logger } = require('~/config');
 
 /**
  * Loads the default models for the application.
@@ -15,31 +15,68 @@ const {
  * @param {Express.Request} req - The Express request object.
  */
 async function loadDefaultModels(req) {
-  const google = getGoogleModels();
-  const openAI = await getOpenAIModels({ user: req.user.id });
-  const anthropic = getAnthropicModels();
-  const chatGPTBrowser = getChatGPTBrowserModels();
-  const azureOpenAI = await getOpenAIModels({ user: req.user.id, azure: true });
-  const gptPlugins = await getOpenAIModels({
-    user: req.user.id,
-    azure: useAzurePlugins,
-    plugins: true,
-  });
-  const assistants = await getOpenAIModels({ assistants: true });
-  const azureAssistants = await getOpenAIModels({ azureAssistants: true });
+  try {
+    const [
+      openAI,
+      anthropic,
+      azureOpenAI,
+      gptPlugins,
+      assistants,
+      azureAssistants,
+      google,
+      bedrock,
+    ] = await Promise.all([
+      getOpenAIModels({ user: req.user.id }).catch((error) => {
+        logger.error('Error fetching OpenAI models:', error);
+        return [];
+      }),
+      getAnthropicModels({ user: req.user.id }).catch((error) => {
+        logger.error('Error fetching Anthropic models:', error);
+        return [];
+      }),
+      getOpenAIModels({ user: req.user.id, azure: true }).catch((error) => {
+        logger.error('Error fetching Azure OpenAI models:', error);
+        return [];
+      }),
+      getOpenAIModels({ user: req.user.id, azure: useAzurePlugins, plugins: true }).catch(
+        (error) => {
+          logger.error('Error fetching Plugin models:', error);
+          return [];
+        },
+      ),
+      getOpenAIModels({ assistants: true }).catch((error) => {
+        logger.error('Error fetching OpenAI Assistants API models:', error);
+        return [];
+      }),
+      getOpenAIModels({ azureAssistants: true }).catch((error) => {
+        logger.error('Error fetching Azure OpenAI Assistants API models:', error);
+        return [];
+      }),
+      Promise.resolve(getGoogleModels()).catch((error) => {
+        logger.error('Error getting Google models:', error);
+        return [];
+      }),
+      Promise.resolve(getBedrockModels()).catch((error) => {
+        logger.error('Error getting Bedrock models:', error);
+        return [];
+      }),
+    ]);
 
-  return {
-    [EModelEndpoint.openAI]: openAI,
-    [EModelEndpoint.agents]: openAI,
-    [EModelEndpoint.google]: google,
-    [EModelEndpoint.anthropic]: anthropic,
-    [EModelEndpoint.gptPlugins]: gptPlugins,
-    [EModelEndpoint.azureOpenAI]: azureOpenAI,
-    [EModelEndpoint.chatGPTBrowser]: chatGPTBrowser,
-    [EModelEndpoint.assistants]: assistants,
-    [EModelEndpoint.azureAssistants]: azureAssistants,
-    [EModelEndpoint.bedrock]: getBedrockModels(),
-  };
+    return {
+      [EModelEndpoint.openAI]: openAI,
+      [EModelEndpoint.agents]: openAI,
+      [EModelEndpoint.google]: google,
+      [EModelEndpoint.anthropic]: anthropic,
+      [EModelEndpoint.gptPlugins]: gptPlugins,
+      [EModelEndpoint.azureOpenAI]: azureOpenAI,
+      [EModelEndpoint.assistants]: assistants,
+      [EModelEndpoint.azureAssistants]: azureAssistants,
+      [EModelEndpoint.bedrock]: bedrock,
+    };
+  } catch (error) {
+    logger.error('Error fetching default models:', error);
+    throw new Error(`Failed to load default models: ${error.message}`);
+  }
 }
 
 module.exports = loadDefaultModels;

api/server/services/Endpoints/agents/initialize.js

@@ -22,6 +22,7 @@ const { getAgent } = require('~/models/Agent');
 const { logger } = require('~/config');
 
 const providerConfigMap = {
+  [Providers.XAI]: initCustom,
   [Providers.OLLAMA]: initCustom,
   [Providers.DEEPSEEK]: initCustom,
   [Providers.OPENROUTER]: initCustom,
@@ -101,6 +102,7 @@ const initializeAgentOptions = async ({
   });
 
   const provider = agent.provider;
+  agent.endpoint = provider;
   let getOptions = providerConfigMap[provider];
   if (!getOptions && providerConfigMap[provider.toLowerCase()] != null) {
     agent.provider = provider.toLowerCase();
@@ -112,9 +114,7 @@ const initializeAgentOptions = async ({
     }
     getOptions = initCustom;
     agent.provider = Providers.OPENAI;
-    agent.endpoint = provider.toLowerCase();
   }
-
   const model_parameters = Object.assign(
     {},
     agent.model_parameters ?? { model: agent.model },

api/server/services/Endpoints/agents/title.js

@@ -20,10 +20,19 @@ const addTitle = async (req, { text, response, client }) => {
 
   const titleCache = getLogStores(CacheKeys.GEN_TITLE);
   const key = `${req.user.id}-${response.conversationId}`;
+  const responseText =
+    response?.content && Array.isArray(response?.content)
+      ? response.content.reduce((acc, block) => {
+        if (block?.type === 'text') {
+          return acc + block.text;
+        }
+        return acc;
+      }, '')
+      : (response?.content ?? response?.text ?? '');
 
   const title = await client.titleConvo({
     text,
-    responseText: response?.text ?? '',
+    responseText,
     conversationId: response.conversationId,
   });
   await titleCache.set(key, title, 120000);

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

@@ -48,7 +48,8 @@ function getClaudeHeaders(model, supportsCacheControl) {
     };
   } else if (/claude-3[-.]7/.test(model)) {
     return {
-      'anthropic-beta': 'output-128k-2025-02-19,prompt-caching-2024-07-31',
+      'anthropic-beta':
+        'token-efficient-tools-2025-02-19,output-128k-2025-02-19,prompt-caching-2024-07-31',
     };
   } else {
     return {

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

@@ -27,6 +27,7 @@ const initializeClient = async ({ req, res, endpointOption, overrideModel, optio
 
   if (anthropicConfig) {
     clientOptions.streamRate = anthropicConfig.streamRate;
+    clientOptions.titleModel = anthropicConfig.titleModel;
   }
 
   /** @type {undefined | TBaseEndpoint} */

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

@@ -1,6 +1,6 @@
 const { HttpsProxyAgent } = require('https-proxy-agent');
 const { anthropicSettings, removeNullishValues } = require('librechat-data-provider');
-const { checkPromptCacheSupport, getClaudeHeaders } = require('./helpers');
+const { checkPromptCacheSupport, getClaudeHeaders, configureReasoning } = require('./helpers');
 
 /**
  * Generates configuration options for creating an Anthropic language model (LLM) instance.
@@ -49,13 +49,14 @@ function getLLMConfig(apiKey, options = {}) {
     clientOptions: {},
   };
 
+  requestOptions = configureReasoning(requestOptions, systemOptions);
+
   if (!/claude-3[-.]7/.test(mergedOptions.model)) {
-    if (mergedOptions.topP !== undefined) {
-      requestOptions.topP = mergedOptions.topP;
-    }
-    if (mergedOptions.topK !== undefined) {
-      requestOptions.topK = mergedOptions.topK;
-    }
+    requestOptions.topP = mergedOptions.topP;
+    requestOptions.topK = mergedOptions.topK;
+  } else if (requestOptions.thinking == null) {
+    requestOptions.topP = mergedOptions.topP;
+    requestOptions.topK = mergedOptions.topK;
   }
 
   const supportsCacheControl =

api/server/services/Endpoints/anthropic/llm.spec.js

@@ -109,4 +109,45 @@ describe('getLLMConfig', () => {
     // Just verifying that the promptCache setting is processed
     expect(result.llmConfig).toBeDefined();
   });
+
+  it('should include topK and topP for Claude-3.7 models when thinking is not enabled', () => {
+    // Test with thinking explicitly set to null/undefined
+    const result = getLLMConfig('test-api-key', {
+      modelOptions: {
+        model: 'claude-3-7-sonnet',
+        topK: 10,
+        topP: 0.9,
+        thinking: false,
+      },
+    });
+
+    expect(result.llmConfig).toHaveProperty('topK', 10);
+    expect(result.llmConfig).toHaveProperty('topP', 0.9);
+
+    // Test with thinking explicitly set to false
+    const result2 = getLLMConfig('test-api-key', {
+      modelOptions: {
+        model: 'claude-3-7-sonnet',
+        topK: 10,
+        topP: 0.9,
+        thinking: false,
+      },
+    });
+
+    expect(result2.llmConfig).toHaveProperty('topK', 10);
+    expect(result2.llmConfig).toHaveProperty('topP', 0.9);
+
+    // Test with decimal notation as well
+    const result3 = getLLMConfig('test-api-key', {
+      modelOptions: {
+        model: 'claude-3.7-sonnet',
+        topK: 10,
+        topP: 0.9,
+        thinking: false,
+      },
+    });
+
+    expect(result3.llmConfig).toHaveProperty('topK', 10);
+    expect(result3.llmConfig).toHaveProperty('topP', 0.9);
+  });
 });

api/server/services/Endpoints/bedrock/build.js

@@ -1,6 +1,5 @@
-const { removeNullishValues, bedrockInputParser } = require('librechat-data-provider');
+const { removeNullishValues } = require('librechat-data-provider');
 const generateArtifactsPrompt = require('~/app/clients/prompts/artifacts');
-const { logger } = require('~/config');
 
 const buildOptions = (endpoint, parsedBody) => {
   const {
@@ -15,12 +14,6 @@ const buildOptions = (endpoint, parsedBody) => {
     artifacts,
     ...model_parameters
   } = parsedBody;
-  let parsedParams = model_parameters;
-  try {
-    parsedParams = bedrockInputParser.parse(model_parameters);
-  } catch (error) {
-    logger.warn('Failed to parse bedrock input', error);
-  }
   const endpointOption = removeNullishValues({
     endpoint,
     name,
@@ -31,7 +24,7 @@ const buildOptions = (endpoint, parsedBody) => {
     spec,
     promptPrefix,
     maxContextTokens,
-    model_parameters: parsedParams,
+    model_parameters,
   });
 
   if (typeof artifacts === 'string') {

api/server/services/Endpoints/bedrock/options.js

@@ -1,14 +1,16 @@
 const { HttpsProxyAgent } = require('https-proxy-agent');
 const {
-  EModelEndpoint,
-  Constants,
   AuthType,
+  Constants,
+  EModelEndpoint,
+  bedrockInputParser,
+  bedrockOutputParser,
   removeNullishValues,
 } = require('librechat-data-provider');
 const { getUserKey, checkUserKeyExpiry } = require('~/server/services/UserService');
 const { sleep } = require('~/server/utils');
 
-const getOptions = async ({ req, endpointOption }) => {
+const getOptions = async ({ req, overrideModel, endpointOption }) => {
   const {
     BEDROCK_AWS_SECRET_ACCESS_KEY,
     BEDROCK_AWS_ACCESS_KEY_ID,
@@ -62,39 +64,44 @@ const getOptions = async ({ req, endpointOption }) => {
 
   /** @type {BedrockClientOptions} */
   const requestOptions = {
-    model: endpointOption.model,
+    model: overrideModel ?? endpointOption.model,
     region: BEDROCK_AWS_DEFAULT_REGION,
-    streaming: true,
-    streamUsage: true,
-    callbacks: [
-      {
-        handleLLMNewToken: async () => {
-          if (!streamRate) {
-            return;
-          }
-          await sleep(streamRate);
-        },
-      },
-    ],
   };
 
-  if (credentials) {
-    requestOptions.credentials = credentials;
-  }
-
-  if (BEDROCK_REVERSE_PROXY) {
-    requestOptions.endpointHost = BEDROCK_REVERSE_PROXY;
-  }
-
   const configOptions = {};
   if (PROXY) {
     /** NOTE: NOT SUPPORTED BY BEDROCK */
     configOptions.httpAgent = new HttpsProxyAgent(PROXY);
   }
 
+  const llmConfig = bedrockOutputParser(
+    bedrockInputParser.parse(
+      removeNullishValues(Object.assign(requestOptions, endpointOption.model_parameters)),
+    ),
+  );
+
+  if (credentials) {
+    llmConfig.credentials = credentials;
+  }
+
+  if (BEDROCK_REVERSE_PROXY) {
+    llmConfig.endpointHost = BEDROCK_REVERSE_PROXY;
+  }
+
+  llmConfig.callbacks = [
+    {
+      handleLLMNewToken: async () => {
+        if (!streamRate) {
+          return;
+        }
+        await sleep(streamRate);
+      },
+    },
+  ];
+
   return {
     /** @type {BedrockClientOptions} */
-    llmConfig: removeNullishValues(Object.assign(requestOptions, endpointOption.model_parameters)),
+    llmConfig,
     configOptions,
   };
 };

api/server/services/Endpoints/custom/initialize.js

@@ -141,7 +141,8 @@ const initializeClient = async ({ req, res, endpointOption, optionsOnly, overrid
         },
         clientOptions,
       );
-      const options = getLLMConfig(apiKey, clientOptions);
+      clientOptions.modelOptions.user = req.user.id;
+      const options = getLLMConfig(apiKey, clientOptions, endpoint);
       if (!customOptions.streamRate) {
         return options;
       }

api/server/services/Endpoints/google/initialize.js

@@ -5,12 +5,7 @@ const { isEnabled } = require('~/server/utils');
 const { GoogleClient } = require('~/app');
 
 const initializeClient = async ({ req, res, endpointOption, overrideModel, optionsOnly }) => {
-  const {
-    GOOGLE_KEY,
-    GOOGLE_REVERSE_PROXY,
-    GOOGLE_AUTH_HEADER,
-    PROXY,
-  } = process.env;
+  const { GOOGLE_KEY, GOOGLE_REVERSE_PROXY, GOOGLE_AUTH_HEADER, PROXY } = process.env;
   const isUserProvided = GOOGLE_KEY === 'user_provided';
   const { key: expiresAt } = req.body;
 
@@ -43,6 +38,7 @@ const initializeClient = async ({ req, res, endpointOption, overrideModel, optio
 
   if (googleConfig) {
     clientOptions.streamRate = googleConfig.streamRate;
+    clientOptions.titleModel = googleConfig.titleModel;
   }
 
   if (allConfig) {

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

@@ -113,6 +113,7 @@ const initializeClient = async ({
 
   if (!isAzureOpenAI && openAIConfig) {
     clientOptions.streamRate = openAIConfig.streamRate;
+    clientOptions.titleModel = openAIConfig.titleModel;
   }
 
   /** @type {undefined | TBaseEndpoint} */
@@ -140,6 +141,7 @@ const initializeClient = async ({
       },
       clientOptions,
     );
+    clientOptions.modelOptions.user = req.user.id;
     const options = getLLMConfig(apiKey, clientOptions);
     if (!clientOptions.streamRate) {
       return options;

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

@@ -9,6 +9,7 @@ const { isEnabled } = require('~/server/utils');
  * @param {Object} options - Additional options for configuring the LLM.
  * @param {Object} [options.modelOptions] - Model-specific options.
  * @param {string} [options.modelOptions.model] - The name of the model to use.
+ * @param {string} [options.modelOptions.user] - The user ID
  * @param {number} [options.modelOptions.temperature] - Controls randomness in output generation (0-2).
  * @param {number} [options.modelOptions.top_p] - Controls diversity via nucleus sampling (0-1).
  * @param {number} [options.modelOptions.frequency_penalty] - Reduces repetition of token sequences (-2 to 2).
@@ -23,13 +24,13 @@ const { isEnabled } = require('~/server/utils');
  * @param {boolean} [options.streaming] - Whether to use streaming mode.
  * @param {Object} [options.addParams] - Additional parameters to add to the model options.
  * @param {string[]} [options.dropParams] - Parameters to remove from the model options.
+ * @param {string|null} [endpoint=null] - The endpoint name
  * @returns {Object} Configuration options for creating an LLM instance.
  */
-function getLLMConfig(apiKey, options = {}) {
+function getLLMConfig(apiKey, options = {}, endpoint = null) {
   const {
     modelOptions = {},
     reverseProxyUrl,
-    useOpenRouter,
     defaultQuery,
     headers,
     proxy,
@@ -56,9 +57,14 @@ function getLLMConfig(apiKey, options = {}) {
     });
   }
 
+  let useOpenRouter;
   /** @type {OpenAIClientOptions['configuration']} */
   const configOptions = {};
-  if (useOpenRouter || (reverseProxyUrl && reverseProxyUrl.includes(KnownEndpoints.openrouter))) {
+  if (
+    (reverseProxyUrl && reverseProxyUrl.includes(KnownEndpoints.openrouter)) ||
+    (endpoint && endpoint.toLowerCase().includes(KnownEndpoints.openrouter))
+  ) {
+    useOpenRouter = true;
     llmConfig.include_reasoning = true;
     configOptions.baseURL = reverseProxyUrl;
     configOptions.defaultHeaders = Object.assign(
@@ -118,6 +124,13 @@ function getLLMConfig(apiKey, options = {}) {
     llmConfig.organization = process.env.OPENAI_ORGANIZATION;
   }
 
+  if (useOpenRouter && llmConfig.reasoning_effort != null) {
+    llmConfig.reasoning = {
+      effort: llmConfig.reasoning_effort,
+    };
+    delete llmConfig.reasoning_effort;
+  }
+
   return {
     /** @type {OpenAIClientOptions} */
     llmConfig,

api/server/services/Files/Code/crud.js

@@ -1,4 +1,3 @@
-// Code Files
 const axios = require('axios');
 const FormData = require('form-data');
 const { getCodeBaseURL } = require('@librechat/agents');
@@ -16,7 +15,8 @@ const MAX_FILE_SIZE = 150 * 1024 * 1024;
 async function getCodeOutputDownloadStream(fileIdentifier, apiKey) {
   try {
     const baseURL = getCodeBaseURL();
-    const response = await axios({
+    /** @type {import('axios').AxiosRequestConfig} */
+    const options = {
       method: 'get',
       url: `${baseURL}/download/${fileIdentifier}`,
       responseType: 'stream',
@@ -25,10 +25,22 @@ async function getCodeOutputDownloadStream(fileIdentifier, apiKey) {
         'X-API-Key': apiKey,
       },
       timeout: 15000,
-    });
+    };
 
+    if (process.env.PROXY) {
+      options.proxy = {
+        host: process.env.PROXY,
+        protocol: process.env.PROXY.startsWith('https') ? 'https' : 'http',
+      };
+    }
+
+    const response = await axios(options);
     return response;
   } catch (error) {
+    logAxiosError({
+      message: `Error downloading code environment file stream: ${error.message}`,
+      error,
+    });
     throw new Error(`Error downloading file: ${error.message}`);
   }
 }
@@ -54,7 +66,8 @@ async function uploadCodeEnvFile({ req, stream, filename, apiKey, entity_id = ''
     form.append('file', stream, filename);
 
     const baseURL = getCodeBaseURL();
-    const response = await axios.post(`${baseURL}/upload`, form, {
+    /** @type {import('axios').AxiosRequestConfig} */
+    const options = {
       headers: {
         ...form.getHeaders(),
         'Content-Type': 'multipart/form-data',
@@ -64,7 +77,16 @@ async function uploadCodeEnvFile({ req, stream, filename, apiKey, entity_id = ''
       },
       maxContentLength: MAX_FILE_SIZE,
       maxBodyLength: MAX_FILE_SIZE,
-    });
+    };
+
+    if (process.env.PROXY) {
+      options.proxy = {
+        host: process.env.PROXY,
+        protocol: process.env.PROXY.startsWith('https') ? 'https' : 'http',
+      };
+    }
+
+    const response = await axios.post(`${baseURL}/upload`, form, options);
 
     /** @type {{ message: string; session_id: string; files: Array<{ fileId: string; filename: string }> }} */
     const result = response.data;

api/server/services/ModelService.js

@@ -4,7 +4,9 @@ const { HttpsProxyAgent } = require('https-proxy-agent');
 const { EModelEndpoint, defaultModels, CacheKeys } = require('librechat-data-provider');
 const { inputSchema, logAxiosError, extractBaseURL, processModelData } = require('~/utils');
 const { OllamaClient } = require('~/app/clients/OllamaClient');
+const { isUserProvided } = require('~/server/utils');
 const getLogStores = require('~/cache/getLogStores');
+const { logger } = require('~/config');
 
 /**
  * Splits a string by commas and trims each resulting value.
@@ -42,7 +44,7 @@ const fetchModels = async ({
   user,
   apiKey,
   baseURL,
-  name = 'OpenAI',
+  name = EModelEndpoint.openAI,
   azure = false,
   userIdQuery = false,
   createTokenConfig = true,
@@ -64,12 +66,19 @@ const fetchModels = async ({
 
   try {
     const options = {
-      headers: {
-        Authorization: `Bearer ${apiKey}`,
-      },
+      headers: {},
       timeout: 5000,
     };
 
+    if (name === EModelEndpoint.anthropic) {
+      options.headers = {
+        'x-api-key': apiKey,
+        'anthropic-version': process.env.ANTHROPIC_VERSION || '2023-06-01',
+      };
+    } else {
+      options.headers.Authorization = `Bearer ${apiKey}`;
+    }
+
     if (process.env.PROXY) {
       options.httpsAgent = new HttpsProxyAgent(process.env.PROXY);
     }
@@ -148,7 +157,7 @@ const fetchOpenAIModels = async (opts, _models = []) => {
       baseURL,
       azure: opts.azure,
       user: opts.user,
-      name: baseURL,
+      name: EModelEndpoint.openAI,
     });
   }
 
@@ -157,7 +166,7 @@ const fetchOpenAIModels = async (opts, _models = []) => {
   }
 
   if (baseURL === openaiBaseURL) {
-    const regex = /(text-davinci-003|gpt-|o\d+-)/;
+    const regex = /(text-davinci-003|gpt-|o\d+)/;
     const excludeRegex = /audio|realtime/;
     models = models.filter((model) => regex.test(model) && !excludeRegex.test(model));
     const instructModels = models.filter((model) => model.includes('instruct'));
@@ -231,13 +240,71 @@ const getChatGPTBrowserModels = () => {
   return models;
 };
 
-const getAnthropicModels = () => {
+/**
+ * Fetches models from the Anthropic API.
+ * @async
+ * @function
+ * @param {object} opts - The options for fetching the models.
+ * @param {string} opts.user - The user ID to send to the API.
+ * @param {string[]} [_models=[]] - The models to use as a fallback.
+ */
+const fetchAnthropicModels = async (opts, _models = []) => {
+  let models = _models.slice() ?? [];
+  let apiKey = process.env.ANTHROPIC_API_KEY;
+  const anthropicBaseURL = 'https://api.anthropic.com/v1';
+  let baseURL = anthropicBaseURL;
+  let reverseProxyUrl = process.env.ANTHROPIC_REVERSE_PROXY;
+
+  if (reverseProxyUrl) {
+    baseURL = extractBaseURL(reverseProxyUrl);
+  }
+
+  if (!apiKey) {
+    return models;
+  }
+
+  const modelsCache = getLogStores(CacheKeys.MODEL_QUERIES);
+
+  const cachedModels = await modelsCache.get(baseURL);
+  if (cachedModels) {
+    return cachedModels;
+  }
+
+  if (baseURL) {
+    models = await fetchModels({
+      apiKey,
+      baseURL,
+      user: opts.user,
+      name: EModelEndpoint.anthropic,
+      tokenKey: EModelEndpoint.anthropic,
+    });
+  }
+
+  if (models.length === 0) {
+    return _models;
+  }
+
+  await modelsCache.set(baseURL, models);
+  return models;
+};
+
+const getAnthropicModels = async (opts = {}) => {
   let models = defaultModels[EModelEndpoint.anthropic];
   if (process.env.ANTHROPIC_MODELS) {
     models = splitAndTrim(process.env.ANTHROPIC_MODELS);
+    return models;
   }
 
-  return models;
+  if (isUserProvided(process.env.ANTHROPIC_API_KEY)) {
+    return models;
+  }
+
+  try {
+    return await fetchAnthropicModels(opts, models);
+  } catch (error) {
+    logger.error('Error fetching Anthropic models:', error);
+    return models;
+  }
 };
 
 const getGoogleModels = () => {

api/server/services/ModelService.spec.js

@@ -352,15 +352,15 @@ describe('splitAndTrim', () => {
 });
 
 describe('getAnthropicModels', () => {
-  it('returns default models when ANTHROPIC_MODELS is not set', () => {
+  it('returns default models when ANTHROPIC_MODELS is not set', async () => {
     delete process.env.ANTHROPIC_MODELS;
-    const models = getAnthropicModels();
+    const models = await getAnthropicModels();
     expect(models).toEqual(defaultModels[EModelEndpoint.anthropic]);
   });
 
-  it('returns models from ANTHROPIC_MODELS when set', () => {
+  it('returns models from ANTHROPIC_MODELS when set', async () => {
     process.env.ANTHROPIC_MODELS = 'claude-1, claude-2 ';
-    const models = getAnthropicModels();
+    const models = await getAnthropicModels();
     expect(models).toEqual(['claude-1', 'claude-2']);
   });
 });