🤖 feat: Custom Endpoint Agents (experimental) (#4627)

* wip: first pass, custom endpoint agents

* chore: imports

* chore: consolidate exports

* fix: imports

* feat: convert message.content array to strings for legacy format handling (deepseek/groq)

* refactor: normalize ollama endpoint name

* refactor: update mocking in isDomainAllowed.spec.js

* refactor: update deepseekModels in tokens.js and tokens.spec.js

api/server/services/Config/getCustomConfig.js

@@ -1,4 +1,4 @@
-const { CacheKeys } = require('librechat-data-provider');
+const { CacheKeys, EModelEndpoint } = require('librechat-data-provider');
 const loadCustomConfig = require('./loadCustomConfig');
 const getLogStores = require('~/cache/getLogStores');
 
@@ -22,4 +22,19 @@ async function getCustomConfig() {
   return customConfig;
 }
 
-module.exports = getCustomConfig;
+/**
+ *
+ * @param {string | EModelEndpoint} endpoint
+ */
+const getCustomEndpointConfig = async (endpoint) => {
+  const customConfig = await getCustomConfig();
+  if (!customConfig) {
+    throw new Error(`Config not found for the ${endpoint} custom endpoint.`);
+  }
+
+  const { endpoints = {} } = customConfig;
+  const customEndpoints = endpoints[EModelEndpoint.custom] ?? [];
+  return customEndpoints.find((endpointConfig) => endpointConfig.name === endpoint);
+};
+
+module.exports = { getCustomConfig, getCustomEndpointConfig };

api/server/services/Config/index.js

@@ -10,12 +10,12 @@ const loadDefaultEndpointsConfig = require('./loadDefaultEConfig');
 
 module.exports = {
   config,
-  getCustomConfig,
   loadCustomConfig,
   loadConfigModels,
   loadDefaultModels,
   loadOverrideConfig,
   loadAsyncEndpoints,
+  ...getCustomConfig,
   loadConfigEndpoints,
   loadDefaultEndpointsConfig,
 };

api/server/services/Config/loadConfigEndpoints.js

@@ -1,6 +1,6 @@
 const { EModelEndpoint, extractEnvVariable } = require('librechat-data-provider');
+const { getCustomConfig } = require('./getCustomConfig');
 const { isUserProvided } = require('~/server/utils');
-const getCustomConfig = require('./getCustomConfig');
 
 /**
  * Load config endpoints from the cached configuration object

api/server/services/Config/loadConfigModels.js

@@ -1,7 +1,16 @@
+const { Providers } = require('@librechat/agents');
 const { EModelEndpoint, extractEnvVariable } = require('librechat-data-provider');
 const { fetchModels } = require('~/server/services/ModelService');
+const { getCustomConfig } = require('./getCustomConfig');
 const { isUserProvided } = require('~/server/utils');
-const getCustomConfig = require('./getCustomConfig');
+
+/**
+ * @param {string} name
+ * @returns {string}
+ */
+function normalizeEndpointName(name = '') {
+  return name.toLowerCase() === Providers.OLLAMA ? Providers.OLLAMA : name;
+}
 
 /**
  * Load config endpoints from the cached configuration object
@@ -61,7 +70,8 @@ async function loadConfigModels(req) {
 
   for (let i = 0; i < customEndpoints.length; i++) {
     const endpoint = customEndpoints[i];
-    const { models, name, baseURL, apiKey } = endpoint;
+    const { models, name: configName, baseURL, apiKey } = endpoint;
+    const name = normalizeEndpointName(configName);
     endpointsMap[name] = endpoint;
 
     const API_KEY = extractEnvVariable(apiKey);

api/server/services/Config/loadConfigModels.spec.js

@@ -1,6 +1,6 @@
 const { fetchModels } = require('~/server/services/ModelService');
+const { getCustomConfig } = require('./getCustomConfig');
 const loadConfigModels = require('./loadConfigModels');
-const getCustomConfig = require('./getCustomConfig');
 
 jest.mock('~/server/services/ModelService');
 jest.mock('./getCustomConfig');
@@ -253,13 +253,13 @@ describe('loadConfigModels', () => {
       }),
     );
 
-    // For groq and Ollama, since the apiKey is "user_provided", models should not be fetched
+    // For groq and ollama, since the apiKey is "user_provided", models should not be fetched
     // Depending on your implementation's behavior regarding "default" models without fetching,
     // you may need to adjust the following assertions:
     expect(result.groq).toBe(exampleConfig.endpoints.custom[2].models.default);
-    expect(result.Ollama).toBe(exampleConfig.endpoints.custom[3].models.default);
+    expect(result.ollama).toBe(exampleConfig.endpoints.custom[3].models.default);
 
-    // Verifying fetchModels was not called for groq and Ollama
+    // Verifying fetchModels was not called for groq and ollama
     expect(fetchModels).not.toHaveBeenCalledWith(
       expect.objectContaining({
         name: 'groq',
@@ -267,7 +267,7 @@ describe('loadConfigModels', () => {
     );
     expect(fetchModels).not.toHaveBeenCalledWith(
       expect.objectContaining({
-        name: 'Ollama',
+        name: 'ollama',
       }),
     );
   });
@@ -335,4 +335,68 @@ describe('loadConfigModels', () => {
 
     expect(result.FalsyFetchModel).toEqual(['defaultModel1', 'defaultModel2']);
   });
+
+  it('normalizes Ollama endpoint name to lowercase', async () => {
+    const testCases = [
+      {
+        name: 'Ollama',
+        apiKey: 'user_provided',
+        baseURL: 'http://localhost:11434/v1/',
+        models: {
+          default: ['mistral', 'llama2'],
+          fetch: false,
+        },
+      },
+      {
+        name: 'OLLAMA',
+        apiKey: 'user_provided',
+        baseURL: 'http://localhost:11434/v1/',
+        models: {
+          default: ['mixtral', 'codellama'],
+          fetch: false,
+        },
+      },
+      {
+        name: 'OLLaMA',
+        apiKey: 'user_provided',
+        baseURL: 'http://localhost:11434/v1/',
+        models: {
+          default: ['phi', 'neural-chat'],
+          fetch: false,
+        },
+      },
+    ];
+
+    getCustomConfig.mockResolvedValue({
+      endpoints: {
+        custom: testCases,
+      },
+    });
+
+    const result = await loadConfigModels(mockRequest);
+
+    // All variations of "Ollama" should be normalized to lowercase "ollama"
+    // and the last config in the array should override previous ones
+    expect(result.Ollama).toBeUndefined();
+    expect(result.OLLAMA).toBeUndefined();
+    expect(result.OLLaMA).toBeUndefined();
+    expect(result.ollama).toEqual(['phi', 'neural-chat']);
+
+    // Verify fetchModels was not called since these are user_provided
+    expect(fetchModels).not.toHaveBeenCalledWith(
+      expect.objectContaining({
+        name: 'Ollama',
+      }),
+    );
+    expect(fetchModels).not.toHaveBeenCalledWith(
+      expect.objectContaining({
+        name: 'OLLAMA',
+      }),
+    );
+    expect(fetchModels).not.toHaveBeenCalledWith(
+      expect.objectContaining({
+        name: 'OLLaMA',
+      }),
+    );
+  });
 });