Skip to content

Enhanced ChatGPT Clone: Features Agents, MCP, Skills, DeepSeek, Anthropic, AWS, OpenAI, Responses API, Azure, Groq, o1, GPT-5, Mistral, OpenRouter, Vertex AI, Gemini, Artifacts, AI model switching, message search, Code Interpreter, langchain, DALL-E-3, OpenAPI Actions, Functions, Secure Multi-User Auth, Presets, open-source for self-hosting. Active

🔧 refactor: Improve Params Handling, Remove Legacy Items, & Update Configs (#6074)

Danny Avilabe280004cfa1
diff --git a/.env.example b/.env.example
index 5fb73557db..94a6d80d88 100644
--- a/.env.example
+++ b/.env.example
@@ -209,12 +209,6 @@ ASSISTANTS_API_KEY=user_provided
# More info, including how to enable use of Assistants with Azure here:
# https://www.librechat.ai/docs/configuration/librechat_yaml/ai_endpoints/azure#using-assistants-with-azure
-#============#
-# OpenRouter #
-#============#
-# !!!Warning: Use the variable above instead of this one. Using this one will override the OpenAI endpoint
-# OPENROUTER_API_KEY=
-
#============#
# Plugins #
#============#
diff --git a/api/app/clients/AnthropicClient.js b/api/app/clients/AnthropicClient.js
index 1f7eb90b26..a2ab752bc2 100644
--- a/api/app/clients/AnthropicClient.js
+++ b/api/app/clients/AnthropicClient.js
@@ -7,8 +7,7 @@ const {
getResponseSender,
validateVisionModel,
} = require('librechat-data-provider');
-const { SplitStreamHandler, GraphEvents } = require('@librechat/agents');
-const { encodeAndFormat } = require('~/server/services/Files/images/encode');
+const { SplitStreamHandler: _Handler, GraphEvents } = require('@librechat/agents');
const {
truncateText,
formatMessage,
@@ -24,6 +23,7 @@ const {
} = require('~/server/services/Endpoints/anthropic/helpers');
const { getModelMaxTokens, getModelMaxOutputTokens, matchModelName } = require('~/utils');
const { spendTokens, spendStructuredTokens } = require('~/models/spendTokens');
+const { encodeAndFormat } = require('~/server/services/Files/images/encode');
const Tokenizer = require('~/server/services/Tokenizer');
const { logger, sendEvent } = require('~/config');
const { sleep } = require('~/server/utils');
@@ -32,6 +32,15 @@ const BaseClient = require('./BaseClient');
const HUMAN_PROMPT = '\n\nHuman:';
const AI_PROMPT = '\n\nAssistant:';
+class SplitStreamHandler extends _Handler {
+ getDeltaContent(chunk) {
+ return (chunk?.delta?.text ?? chunk?.completion) || '';
+ }
+ getReasoningDelta(chunk) {
+ return chunk?.delta?.thinking || '';
+ }
+}
+
/** Helper function to introduce a delay before retrying */
function delayBeforeRetry(attempts, baseDelay = 1000) {
return new Promise((resolve) => setTimeout(resolve, baseDelay * attempts));
@@ -105,7 +114,9 @@ class AnthropicClient extends BaseClient {
const modelMatch = matchModelName(this.modelOptions.model, EModelEndpoint.anthropic);
this.isClaude3 = modelMatch.includes('claude-3');
- this.isLegacyOutput = !modelMatch.includes('claude-3-5-sonnet');
+ this.isLegacyOutput = !(
+ /claude-3[-.]5-sonnet/.test(modelMatch) || /claude-3[-.]7/.test(modelMatch)
+ );
this.supportsCacheControl = this.options.promptCache && checkPromptCacheSupport(modelMatch);
if (
@@ -733,10 +744,17 @@ class AnthropicClient extends BaseClient {
stop_sequences,
temperature,
metadata,
- top_p,
- top_k,
};
+ if (!/claude-3[-.]7/.test(model)) {
+ if (top_p !== undefined) {
+ requestOptions.top_p = top_p;
+ }
+ if (top_k !== undefined) {
+ requestOptions.top_k = top_k;
+ }
+ }
+
if (this.useMessages) {
requestOptions.messages = payload;
requestOptions.max_tokens =
@@ -798,50 +816,16 @@ class AnthropicClient extends BaseClient {
}
});
- /** @param {string} chunk */
- const handleChunk = (chunk) => {
- this.streamHandler.handle({
- choices: [
- {
- delta: {
- content: chunk,
- },
- },
- ],
- });
- };
- /** @param {string} chunk */
- const handleReasoningChunk = (chunk) => {
- this.streamHandler.handle({
- choices: [
- {
- delta: {
- reasoning_content: chunk,
- },
- },
- ],
- });
- };
-
for await (const completion of response) {
- // Handle each completion as before
const type = completion?.type ?? '';
if (tokenEventTypes.has(type)) {
logger.debug(`[AnthropicClient] ${type}`, completion);
this[type] = completion;
}
- if (completion?.delta?.thinking) {
- handleReasoningChunk(completion.delta.thinking);
- } else if (completion?.delta?.text) {
- handleChunk(completion.delta.text);
- } else if (completion.completion) {
- handleChunk(completion.completion);
- }
-
+ this.streamHandler.handle(completion);
await sleep(streamRate);
}
- // Successful processing, exit loop
break;
} catch (error) {
attempts += 1;
diff --git a/api/app/clients/BaseClient.js b/api/app/clients/BaseClient.js
index ebf3ca12d9..61b39a8f6d 100644
--- a/api/app/clients/BaseClient.js
+++ b/api/app/clients/BaseClient.js
@@ -5,10 +5,11 @@ const {
isAgentsEndpoint,
isParamEndpoint,
EModelEndpoint,
+ excludedKeys,
ErrorTypes,
Constants,
} = require('librechat-data-provider');
-const { getMessages, saveMessage, updateMessage, saveConvo } = require('~/models');
+const { getMessages, saveMessage, updateMessage, saveConvo, getConvo } = require('~/models');
const { addSpaceIfNeeded, isEnabled } = require('~/server/utils');
const { truncateToolCallOutputs } = require('./prompts');
const checkBalance = require('~/models/checkBalance');
@@ -55,6 +56,10 @@ class BaseClient {
* Flag to determine if the client re-submitted the latest assistant message.
* @type {boolean | undefined} */
this.continued;
+ /**
+ * Flag to determine if the client has already fetched the conversation while saving new messages.
+ * @type {boolean | undefined} */
+ this.fetchedConvo;
/** @type {TMessage[]} */
this.currentMessages = [];
/** @type {import('librechat-data-provider').VisionModes | undefined} */
@@ -863,16 +868,39 @@ class BaseClient {
return { message: savedMessage };
}
- const conversation = await saveConvo(
- this.options.req,
- {
- conversationId: message.conversationId,
- endpoint: this.options.endpoint,
- endpointType: this.options.endpointType,
- ...endpointOptions,
- },
- { context: 'api/app/clients/BaseClient.js - saveMessageToDatabase #saveConvo' },
- );
+ const fieldsToKeep = {
+ conversationId: message.conversationId,
+ endpoint: this.options.endpoint,
+ endpointType: this.options.endpointType,
+ ...endpointOptions,
+ };
+
+ const existingConvo =
+ this.fetchedConvo === true
+ ? null
+ : await getConvo(this.options.req?.user?.id, message.conversationId);
+
+ const unsetFields = {};
+ if (existingConvo != null) {
+ this.fetchedConvo = true;
+ for (const key in existingConvo) {
+ if (!key) {
+ continue;
+ }
+ if (excludedKeys.has(key)) {
+ continue;
+ }
+
+ if (endpointOptions?.[key] === undefined) {
+ unsetFields[key] = 1;
+ }
+ }
+ }
+
+ const conversation = await saveConvo(this.options.req, fieldsToKeep, {
+ context: 'api/app/clients/BaseClient.js - saveMessageToDatabase #saveConvo',
+ unsetFields,
+ });
return { message: savedMessage, conversation };
}
diff --git a/api/app/clients/OpenAIClient.js b/api/app/clients/OpenAIClient.js
index 7bd7879dcf..8d0bce25d2 100644
--- a/api/app/clients/OpenAIClient.js
+++ b/api/app/clients/OpenAIClient.js
@@ -109,12 +109,7 @@ class OpenAIClient extends BaseClient {
const omniPattern = /\b(o1|o3)\b/i;
this.isOmni = omniPattern.test(this.modelOptions.model);
- const { OPENROUTER_API_KEY, OPENAI_FORCE_PROMPT } = process.env ?? {};
- if (OPENROUTER_API_KEY && !this.azure) {
- this.apiKey = OPENROUTER_API_KEY;
- this.useOpenRouter = true;
- }
-
+ const { OPENAI_FORCE_PROMPT } = process.env ?? {};
const { reverseProxyUrl: reverseProxy } = this.options;
if (!this.useOpenRouter && reverseProxy && reverseProxy.includes(KnownEndpoints.openrouter)) {
diff --git a/api/app/clients/specs/AnthropicClient.test.js b/api/app/clients/specs/AnthropicClient.test.js
index eef6bb6748..b565e6d188 100644
--- a/api/app/clients/specs/AnthropicClient.test.js
+++ b/api/app/clients/specs/AnthropicClient.test.js
@@ -1,3 +1,4 @@
+const { SplitStreamHandler } = require('@librechat/agents');
const { anthropicSettings } = require('librechat-data-provider');
const AnthropicClient = require('~/app/clients/AnthropicClient');
@@ -405,4 +406,278 @@ describe('AnthropicClient', () => {
expect(Number.isNaN(result)).toBe(false);
});
});
+
+ describe('maxOutputTokens handling for different models', () => {
+ it('should not cap maxOutputTokens for Claude 3.5 Sonnet models', () => {
+ const client = new AnthropicClient('test-api-key');
+ const highTokenValue = anthropicSettings.legacy.maxOutputTokens.default * 10;
+
+ client.setOptions({
+ modelOptions: {
+ model: 'claude-3-5-sonnet',
+ maxOutputTokens: highTokenValue,
+ },
+ });
+
+ expect(client.modelOptions.maxOutputTokens).toBe(highTokenValue);
+
+ // Test with decimal notation
+ client.setOptions({
+ modelOptions: {
+ model: 'claude-3.5-sonnet',
+ maxOutputTokens: highTokenValue,
+ },
+ });
+
+ expect(client.modelOptions.maxOutputTokens).toBe(highTokenValue);
+ });
+
+ it('should not cap maxOutputTokens for Claude 3.7 models', () => {
+ const client = new AnthropicClient('test-api-key');
+ const highTokenValue = anthropicSettings.legacy.maxOutputTokens.default * 2;
+
+ client.setOptions({
+ modelOptions: {
+ model: 'claude-3-7-sonnet',
+ maxOutputTokens: highTokenValue,
+ },
+ });
+
+ expect(client.modelOptions.maxOutputTokens).toBe(highTokenValue);
+
+ // Test with decimal notation
+ client.setOptions({
+ modelOptions: {
+ model: 'claude-3.7-sonnet',
+ maxOutputTokens: highTokenValue,
+ },
+ });
+
+ expect(client.modelOptions.maxOutputTokens).toBe(highTokenValue);
+ });
+
+ it('should cap maxOutputTokens for Claude 3.5 Haiku models', () => {
+ const client = new AnthropicClient('test-api-key');
+ const highTokenValue = anthropicSettings.legacy.maxOutputTokens.default * 2;
+
+ client.setOptions({
+ modelOptions: {
+ model: 'claude-3-5-haiku',
+ maxOutputTokens: highTokenValue,
+ },
+ });
+
+ expect(client.modelOptions.maxOutputTokens).toBe(
+ anthropicSettings.legacy.maxOutputTokens.default,
+ );
+
+ // Test with decimal notation
+ client.setOptions({
+ modelOptions: {
+ model: 'claude-3.5-haiku',
+ maxOutputTokens: highTokenValue,
+ },
+ });
+
+ expect(client.modelOptions.maxOutputTokens).toBe(
+ anthropicSettings.legacy.maxOutputTokens.default,
+ );
+ });
+
+ it('should cap maxOutputTokens for Claude 3 Haiku and Opus models', () => {
+ const client = new AnthropicClient('test-api-key');
+ const highTokenValue = anthropicSettings.legacy.maxOutputTokens.default * 2;
+
+ // Test haiku
+ client.setOptions({
+ modelOptions: {
+ model: 'claude-3-haiku',
+ maxOutputTokens: highTokenValue,
+ },
+ });
+
+ expect(client.modelOptions.maxOutputTokens).toBe(
+ anthropicSettings.legacy.maxOutputTokens.default,
+ );
+
+ // Test opus
+ client.setOptions({
+ modelOptions: {
+ model: 'claude-3-opus',
+ maxOutputTokens: highTokenValue,
+ },
+ });
+
+ expect(client.modelOptions.maxOutputTokens).toBe(
+ anthropicSettings.legacy.maxOutputTokens.default,
+ );
+ });
+ });
+
+ describe('topK/topP parameters for different models', () => {
+ beforeEach(() => {
+ // Mock the SplitStreamHandler
+ jest.spyOn(SplitStreamHandler.prototype, 'handle').mockImplementation(() => {});
+ });
+
+ afterEach(() => {
+ jest.restoreAllMocks();
+ });
+
+ it('should include top_k and top_p parameters for non-claude-3.7 models', async () => {
+ const client = new AnthropicClient('test-api-key');
+
+ // Create a mock async generator function
+ async function* mockAsyncGenerator() {
+ yield { type: 'message_start', message: { usage: {} } };
+ yield { delta: { text: 'Test response' } };
+ yield { type: 'message_delta', usage: {} };
+ }
+
+ // Mock createResponse to return the async generator
+ jest.spyOn(client, 'createResponse').mockImplementation(() => {
+ return mockAsyncGenerator();
+ });
+
+ client.setOptions({
+ modelOptions: {
+ model: 'claude-3-opus',
+ temperature: 0.7,
+ topK: 10,
+ topP: 0.9,
+ },
+ });
+
+ // Mock getClient to capture the request options
+ let capturedOptions = null;
+ jest.spyOn(client, 'getClient').mockImplementation((options) => {
+ capturedOptions = options;
+ return {};
+ });
+
+ const payload = [{ role: 'user', content: 'Test message' }];
+ await client.sendCompletion(payload, {});
+
+ // Check the options passed to getClient
+ expect(capturedOptions).toHaveProperty('top_k', 10);
+ expect(capturedOptions).toHaveProperty('top_p', 0.9);
+ });
+
+ it('should include top_k and top_p parameters for claude-3-5-sonnet models', async () => {
+ const client = new AnthropicClient('test-api-key');
+
+ // Create a mock async generator function
+ async function* mockAsyncGenerator() {
+ yield { type: 'message_start', message: { usage: {} } };
+ yield { delta: { text: 'Test response' } };
+ yield { type: 'message_delta', usage: {} };
+ }
+
+ // Mock createResponse to return the async generator
+ jest.spyOn(client, 'createResponse').mockImplementation(() => {
+ return mockAsyncGenerator();
+ });
+
+ client.setOptions({
+ modelOptions: {
+ model: 'claude-3-5-sonnet',
+ temperature: 0.7,
+ topK: 10,
+ topP: 0.9,
+ },
+ });
+
+ // Mock getClient to capture the request options
+ let capturedOptions = null;
+ jest.spyOn(client, 'getClient').mockImplementation((options) => {
+ capturedOptions = options;
+ return {};
+ });
+
+ const payload = [{ role: 'user', content: 'Test message' }];
+ await client.sendCompletion(payload, {});
+
+ // Check the options passed to getClient
+ expect(capturedOptions).toHaveProperty('top_k', 10);
+ expect(capturedOptions).toHaveProperty('top_p', 0.9);
+ });
+
+ it('should not include top_k and top_p parameters for claude-3-7-sonnet models', async () => {
+ const client = new AnthropicClient('test-api-key');
+
+ // Create a mock async generator function
+ async function* mockAsyncGenerator() {
+ yield { type: 'message_start', message: { usage: {} } };
+ yield { delta: { text: 'Test response' } };
+ yield { type: 'message_delta', usage: {} };
+ }
+
+ // Mock createResponse to return the async generator
+ jest.spyOn(client, 'createResponse').mockImplementation(() => {
+ return mockAsyncGenerator();
+ });
+
+ client.setOptions({
+ modelOptions: {
+ model: 'claude-3-7-sonnet',
+ temperature: 0.7,
+ topK: 10,
+ topP: 0.9,
+ },
+ });
+
+ // Mock getClient to capture the request options
+ let capturedOptions = null;
+ jest.spyOn(client, 'getClient').mockImplementation((options) => {
+ capturedOptions = options;
+ return {};
+ });
+
+ const payload = [{ role: 'user', content: 'Test message' }];
+ await client.sendCompletion(payload, {});
+
+ // Check the options passed to getClient
+ expect(capturedOptions).not.toHaveProperty('top_k');
+ expect(capturedOptions).not.toHaveProperty('top_p');
+ });
+
+ it('should not include top_k and top_p parameters for models with decimal notation (claude-3.7)', async () => {
+ const client = new AnthropicClient('test-api-key');
+
+ // Create a mock async generator function
+ async function* mockAsyncGenerator() {
+ yield { type: 'message_start', message: { usage: {} } };
+ yield { delta: { text: 'Test response' } };
+ yield { type: 'message_delta', usage: {} };
+ }
+
+ // Mock createResponse to return the async generator
+ jest.spyOn(client, 'createResponse').mockImplementation(() => {
+ return mockAsyncGenerator();
+ });
+
+ client.setOptions({
+ modelOptions: {
+ model: 'claude-3.7-sonnet',
+ temperature: 0.7,
+ topK: 10,
+ topP: 0.9,
+ },
+ });
+
+ // Mock getClient to capture the request options
+ let capturedOptions = null;
+ jest.spyOn(client, 'getClient').mockImplementation((options) => {
+ capturedOptions = options;
+ return {};
+ });
+
+ const payload = [{ role: 'user', content: 'Test message' }];
+ await client.sendCompletion(payload, {});
+
+ // Check the options passed to getClient
+ expect(capturedOptions).not.toHaveProperty('top_k');
+ expect(capturedOptions).not.toHaveProperty('top_p');
+ });
+ });
});
diff --git a/api/app/clients/specs/OpenAIClient.test.js b/api/app/clients/specs/OpenAIClient.test.js
index 2aaec518eb..0e811cf38a 100644
--- a/api/app/clients/specs/OpenAIClient.test.js
+++ b/api/app/clients/specs/OpenAIClient.test.js
@@ -202,14 +202,6 @@ describe('OpenAIClient', () => {
expect(client.modelOptions.temperature).toBe(0.7);
});
- it('should set apiKey and useOpenRouter if OPENROUTER_API_KEY is present', () => {
- process.env.OPENROUTER_API_KEY = 'openrouter-key';
- client.setOptions({});
- expect(client.apiKey).toBe('openrouter-key');
- expect(client.useOpenRouter).toBe(true);
- delete process.env.OPENROUTER_API_KEY; // Cleanup
- });
-
it('should set FORCE_PROMPT based on OPENAI_FORCE_PROMPT or reverseProxyUrl', () => {
process.env.OPENAI_FORCE_PROMPT = 'true';
client.setOptions({});
@@ -534,7 +526,6 @@ describe('OpenAIClient', () => {
afterEach(() => {
delete process.env.AZURE_OPENAI_DEFAULT_MODEL;
delete process.env.AZURE_USE_MODEL_AS_DEPLOYMENT_NAME;
- delete process.env.OPENROUTER_API_KEY;
});
it('should call getCompletion and fetchEventSource when using a text/instruct model', async () => {
diff --git a/api/models/Conversation.js b/api/models/Conversation.js
index d6365e99ce..9e51926ebc 100644
--- a/api/models/Conversation.js
+++ b/api/models/Conversation.js
@@ -104,10 +104,16 @@ module.exports = {
update.expiredAt = null;
}
+ /** @type {{ $set: Partial<TConversation>; $unset?: Record<keyof TConversation, number> }} */
+ const updateOperation = { $set: update };
+ if (metadata && metadata.unsetFields && Object.keys(metadata.unsetFields).length > 0) {
+ updateOperation.$unset = metadata.unsetFields;
+ }
+
/** Note: the resulting Model object is necessary for Meilisearch operations */
const conversation = await Conversation.findOneAndUpdate(
{ conversationId, user: req.user.id },
- update,
+ updateOperation,
{
new: true,
upsert: true,
diff --git a/api/models/Token.js b/api/models/Token.js
index 210666ddd7..0ed18320ae 100644
--- a/api/models/Token.js
+++ b/api/models/Token.js
@@ -13,6 +13,13 @@ const Token = mongoose.model('Token', tokenSchema);
*/
async function fixIndexes() {
try {
+ if (
+ process.env.NODE_ENV === 'CI' ||
+ process.env.NODE_ENV === 'development' ||
+ process.env.NODE_ENV === 'test'
+ ) {
+ return;
+ }
const indexes = await Token.collection.indexes();
logger.debug('Existing Token Indexes:', JSON.stringify(indexes, null, 2));
const unwantedTTLIndexes = indexes.filter(
diff --git a/api/models/schema/convoSchema.js b/api/models/schema/convoSchema.js
index 7d8beed6a6..ae50b7cd22 100644
--- a/api/models/schema/convoSchema.js
+++ b/api/models/schema/convoSchema.js
@@ -20,8 +20,6 @@ const convoSchema = mongoose.Schema(
index: true,
},
messages: [{ type: mongoose.Schema.Types.ObjectId, ref: 'Message' }],
- // google only
- examples: { type: [{ type: mongoose.Schema.Types.Mixed }], default: undefined },
agentOptions: {
type: mongoose.Schema.Types.Mixed,
},
@@ -48,12 +46,12 @@ if (process.env.MEILI_HOST && process.env.MEILI_MASTER_KEY) {
convoSchema.plugin(mongoMeili, {
host: process.env.MEILI_HOST,
apiKey: process.env.MEILI_MASTER_KEY,
- indexName: 'convos', // Will get created automatically if it doesn't exist already
+ /** Note: Will get created automatically if it doesn't exist already */
+ indexName: 'convos',
primaryKey: 'conversationId',
});
}
-// Create TTL index
convoSchema.index({ expiredAt: 1 }, { expireAfterSeconds: 0 });
convoSchema.index({ createdAt: 1, updatedAt: 1 });
convoSchema.index({ conversationId: 1, user: 1 }, { unique: true });
diff --git a/api/models/schema/defaults.js b/api/models/schema/defaults.js
index 8e1f816a15..3b3664df73 100644
--- a/api/models/schema/defaults.js
+++ b/api/models/schema/defaults.js
@@ -1,3 +1,5 @@
+const mongoose = require('mongoose');
+
const conversationPreset = {
// endpoint: [azureOpenAI, openAI, anthropic, chatGPTBrowser]
endpoint: {
@@ -24,6 +26,7 @@ const conversationPreset = {
required: false,
},
// for google only
+ examples: { type: [{ type: mongoose.Schema.Types.Mixed }], default: undefined },
modelLabel: {
type: String,
required: false,
@@ -129,56 +132,6 @@ const conversationPreset = {
},
};
-const agentOptions = {
- model: {
- type: String,
- required: false,
- },
- // for azureOpenAI, openAI only
- chatGptLabel: {
- type: String,
- required: false,
- },
- modelLabel: {
- type: String,
- required: false,
- },
- promptPrefix: {
- type: String,
- required: false,
- },
- temperature: {
- type: Number,
- required: false,
- },
- top_p: {
- type: Number,
- required: false,
- },
- // for google only
- topP: {
- type: Number,
- required: false,
- },
- topK: {
- type: Number,
- required: false,
- },
- maxOutputTokens: {
- type: Number,
- required: false,
- },
- presence_penalty: {
- type: Number,
- required: false,
- },
- frequency_penalty: {
- type: Number,
- required: false,
- },
-};
-
module.exports = {
conversationPreset,
- agentOptions,
};
diff --git a/api/models/schema/presetSchema.js b/api/models/schema/presetSchema.js
index e1c92ab9c0..918e5c4069 100644
--- a/api/models/schema/presetSchema.js
+++ b/api/models/schema/presetSchema.js
@@ -23,8 +23,6 @@ const presetSchema = mongoose.Schema(
order: {
type: Number,
},
- // google only
- examples: [{ type: mongoose.Schema.Types.Mixed }],
...conversationPreset,
agentOptions: {
type: mongoose.Schema.Types.Mixed,
diff --git a/api/server/services/Endpoints/anthropic/llm.js b/api/server/services/Endpoints/anthropic/llm.js
index c68fd4738d..186444cec8 100644
--- a/api/server/services/Endpoints/anthropic/llm.js
+++ b/api/server/services/Endpoints/anthropic/llm.js
@@ -43,14 +43,21 @@ function getLLMConfig(apiKey, options = {}) {
model: mergedOptions.model,
stream: mergedOptions.stream,
temperature: mergedOptions.temperature,
- topP: mergedOptions.topP,
- topK: mergedOptions.topK,
stopSequences: mergedOptions.stop,
maxTokens:
mergedOptions.maxOutputTokens || anthropicSettings.maxOutputTokens.reset(mergedOptions.model),
clientOptions: {},
};
+ if (!/claude-3[-.]7/.test(mergedOptions.model)) {
+ if (mergedOptions.topP !== undefined) {
+ requestOptions.topP = mergedOptions.topP;
+ }
+ if (mergedOptions.topK !== undefined) {
+ requestOptions.topK = mergedOptions.topK;
+ }
+ }
+
const supportsCacheControl =
systemOptions.promptCache === true && checkPromptCacheSupport(requestOptions.model);
const headers = getClaudeHeaders(requestOptions.model, supportsCacheControl);
diff --git a/api/server/services/Endpoints/anthropic/llm.spec.js b/api/server/services/Endpoints/anthropic/llm.spec.js
new file mode 100644
index 0000000000..a1dc6a44b6
--- /dev/null
+++ b/api/server/services/Endpoints/anthropic/llm.spec.js
@@ -0,0 +1,112 @@
+const { anthropicSettings } = require('librechat-data-provider');
+const { getLLMConfig } = require('~/server/services/Endpoints/anthropic/llm');
+
+jest.mock('https-proxy-agent', () => ({
+ HttpsProxyAgent: jest.fn().mockImplementation((proxy) => ({ proxy })),
+}));
+
+describe('getLLMConfig', () => {
+ it('should create a basic configuration with default values', () => {
+ const result = getLLMConfig('test-api-key', { modelOptions: {} });
+
+ expect(result.llmConfig).toHaveProperty('apiKey', 'test-api-key');
+ expect(result.llmConfig).toHaveProperty('model', anthropicSettings.model.default);
+ expect(result.llmConfig).toHaveProperty('stream', true);
+ expect(result.llmConfig).toHaveProperty('maxTokens');
+ });
+
+ it('should include proxy settings when provided', () => {
+ const result = getLLMConfig('test-api-key', {
+ modelOptions: {},
+ proxy: 'http://proxy:8080',
+ });
+
+ expect(result.llmConfig.clientOptions).toHaveProperty('httpAgent');
+ expect(result.llmConfig.clientOptions.httpAgent).toHaveProperty('proxy', 'http://proxy:8080');
+ });
+
+ it('should include reverse proxy URL when provided', () => {
+ const result = getLLMConfig('test-api-key', {
+ modelOptions: {},
+ reverseProxyUrl: 'http://reverse-proxy',
+ });
+
+ expect(result.llmConfig.clientOptions).toHaveProperty('baseURL', 'http://reverse-proxy');
+ });
+
+ it('should include topK and topP for non-Claude-3.7 models', () => {
+ const result = getLLMConfig('test-api-key', {
+ modelOptions: {
+ model: 'claude-3-opus',
+ topK: 10,
+ topP: 0.9,
+ },
+ });
+
+ expect(result.llmConfig).toHaveProperty('topK', 10);
+ expect(result.llmConfig).toHaveProperty('topP', 0.9);
+ });
+
+ it('should include topK and topP for Claude-3.5 models', () => {
+ const result = getLLMConfig('test-api-key', {
+ modelOptions: {
+ model: 'claude-3-5-sonnet',
+ topK: 10,
+ topP: 0.9,
+ },
+ });
+
+ expect(result.llmConfig).toHaveProperty('topK', 10);
+ expect(result.llmConfig).toHaveProperty('topP', 0.9);
+ });
+
+ it('should NOT include topK and topP for Claude-3-7 models (hyphen notation)', () => {
+ const result = getLLMConfig('test-api-key', {
+ modelOptions: {
+ model: 'claude-3-7-sonnet',
+ topK: 10,
+ topP: 0.9,
+ },
+ });
+
+ expect(result.llmConfig).not.toHaveProperty('topK');
+ expect(result.llmConfig).not.toHaveProperty('topP');
+ });
+
+ it('should NOT include topK and topP for Claude-3.7 models (decimal notation)', () => {
+ const result = getLLMConfig('test-api-key', {
+ modelOptions: {
+ model: 'claude-3.7-sonnet',
+ topK: 10,
+ topP: 0.9,
+ },
+ });
+
+ expect(result.llmConfig).not.toHaveProperty('topK');
+ expect(result.llmConfig).not.toHaveProperty('topP');
+ });
+
+ it('should handle custom maxOutputTokens', () => {
+ const result = getLLMConfig('test-api-key', {
+ modelOptions: {
+ model: 'claude-3-opus',
+ maxOutputTokens: 2048,
+ },
+ });
+
+ expect(result.llmConfig).toHaveProperty('maxTokens', 2048);
+ });
+
+ it('should handle promptCache setting', () => {
+ const result = getLLMConfig('test-api-key', {
+ modelOptions: {
+ model: 'claude-3-5-sonnet',
+ promptCache: true,
+ },
+ });
+
+ // We're not checking specific header values since that depends on the actual helper function
+ // Just verifying that the promptCache setting is processed
+ expect(result.llmConfig).toBeDefined();
+ });
+});
diff --git a/api/server/services/ModelService.js b/api/server/services/ModelService.js
index 9630f0bd87..48ae85b663 100644
--- a/api/server/services/ModelService.js
+++ b/api/server/services/ModelService.js
@@ -129,9 +129,6 @@ const fetchOpenAIModels = async (opts, _models = []) => {
// .split('/deployments')[0]
// .concat(`/models?api-version=${azure.azureOpenAIApiVersion}`);
// apiKey = azureOpenAIApiKey;
- } else if (process.env.OPENROUTER_API_KEY) {
- reverseProxyUrl = 'https://openrouter.ai/api/v1';
- apiKey = process.env.OPENROUTER_API_KEY;
}
if (reverseProxyUrl) {
@@ -218,7 +215,7 @@ const getOpenAIModels = async (opts) => {
return models;
}
- if (userProvidedOpenAI && !process.env.OPENROUTER_API_KEY) {
+ if (userProvidedOpenAI) {
return models;
}
diff --git a/api/server/services/ModelService.spec.js b/api/server/services/ModelService.spec.js
index a383db1e3c..1fbe347a00 100644
--- a/api/server/services/ModelService.spec.js
+++ b/api/server/services/ModelService.spec.js
@@ -161,22 +161,6 @@ describe('getOpenAIModels', () => {
expect(models).toEqual(expect.arrayContaining(['openai-model', 'openai-model-2']));
});
- it('attempts to use OPENROUTER_API_KEY if set', async () => {
- process.env.OPENROUTER_API_KEY = 'test-router-key';
- const expectedModels = ['model-router-1', 'model-router-2'];
-
- axios.get.mockResolvedValue({
- data: {
- data: expectedModels.map((id) => ({ id })),
- },
- });
-
- const models = await getOpenAIModels({ user: 'user456' });
-
- expect(models).toEqual(expect.arrayContaining(expectedModels));
- expect(axios.get).toHaveBeenCalled();
- });
-
it('utilizes proxy configuration when PROXY is set', async () => {
axios.get.mockResolvedValue({
data: {
diff --git a/client/src/components/SidePanel/Parameters/Panel.tsx b/client/src/components/SidePanel/Parameters/Panel.tsx
index 0b9cc36460..f7e1f579ea 100644
--- a/client/src/components/SidePanel/Parameters/Panel.tsx
+++ b/client/src/components/SidePanel/Parameters/Panel.tsx
@@ -1,5 +1,6 @@
+import { RotateCcw } from 'lucide-react';
import React, { useMemo, useState, useEffect, useCallback } from 'react';
-import { getSettingsKeys, tConvoUpdateSchema } from 'librechat-data-provider';
+import { excludedKeys, getSettingsKeys, tConvoUpdateSchema } from 'librechat-data-provider';
import type { TPreset } from 'librechat-data-provider';
import { SaveAsPresetDialog } from '~/components/Endpoints';
import { useSetIndexOptions, useLocalize } from '~/hooks';
@@ -9,23 +10,6 @@ import { componentMapping } from './components';
import { useChatContext } from '~/Providers';
import { settings } from './settings';
-const excludedKeys = new Set([
- 'conversationId',
- 'title',
- 'endpoint',
- 'endpointType',
- 'createdAt',
- 'updatedAt',
- 'messages',
- 'isArchived',
- 'tags',
- 'user',
- '__v',
- '_id',
- 'tools',
- 'model',
-]);
-
export default function Parameters() {
const localize = useLocalize();
const { conversation, setConversation } = useChatContext();
@@ -105,6 +89,31 @@ export default function Parameters() {
});
}, [parameters, setConversation]);
+ const resetParameters = useCallback(() => {
+ setConversation((prev) => {
+ if (!prev) {
+ return prev;
+ }
+
+ const updatedConversation = { ...prev };
+ const resetKeys: string[] = [];
+
+ Object.keys(updatedConversation).forEach((key) => {
+ if (excludedKeys.has(key)) {
+ return;
+ }
+
+ if (updatedConversation[key] !== undefined) {
+ resetKeys.push(key);
+ delete updatedConversation[key];
+ }
+ });
+
+ logger.log('parameters', 'parameters reset, affected keys:', resetKeys);
+ return updatedConversation;
+ });
+ }, [setConversation]);
+
const openDialog = useCallback(() => {
const newPreset = tConvoUpdateSchema.parse({
...conversation,
@@ -146,7 +155,17 @@ export default function Parameters() {
);
})}
</div>
- <div className="mt-6 flex justify-center">
+ <div className="mt-4 flex justify-center">
+ <button
+ type="button"
+ onClick={resetParameters}
+ className="btn btn-neutral flex w-full items-center justify-center gap-2 px-4 py-2 text-sm"
+ >
+ <RotateCcw className="h-4 w-4" aria-hidden="true" />
+ {localize('com_ui_reset_var', { 0: localize('com_ui_model_parameters') })}
+ </button>
+ </div>
+ <div className="mt-2 flex justify-center">
<button
onClick={openDialog}
className="btn btn-primary focus:shadow-outline flex w-full items-center justify-center px-4 py-2 font-semibold text-white hover:bg-green-600 focus:border-green-500"
diff --git a/client/tsconfig.json b/client/tsconfig.json
index d588c5afea..ba00b0d24c 100644
--- a/client/tsconfig.json
+++ b/client/tsconfig.json
@@ -35,10 +35,5 @@
"test/setupTests.js",
"env.d.ts",
"../config/translations/**/*.ts"
- ],
- "references": [
- {
- "path": "./tsconfig.node.json"
- }
]
}
diff --git a/client/tsconfig.node.json b/client/tsconfig.node.json
deleted file mode 100644
index 9d31e2aed9..0000000000
--- a/client/tsconfig.node.json
+++ /dev/null
@@ -1,9 +0,0 @@
-{
- "compilerOptions": {
- "composite": true,
- "module": "ESNext",
- "moduleResolution": "Node",
- "allowSyntheticDefaultImports": true
- },
- "include": ["vite.config.ts"]
-}
diff --git a/client/vite.config.ts b/client/vite.config.ts
index 94c68bfd54..adac3e70e8 100644
--- a/client/vite.config.ts
+++ b/client/vite.config.ts
@@ -37,13 +37,12 @@ export default defineConfig({
},
useCredentials: true,
workbox: {
- globPatterns: [
- 'assets/**/*.{png,jpg,svg,ico}',
- '**/*.{js,css,html,ico,woff2}',
- ],
+ globPatterns: ['**/*'],
+ globIgnores: ['images/**/*'],
maximumFileSizeToCacheInBytes: 4 * 1024 * 1024,
navigateFallbackDenylist: [/^\/oauth/],
},
+ includeAssets: ['**/*'],
manifest: {
name: 'LibreChat',
short_name: 'LibreChat',
@@ -70,7 +69,7 @@ export default defineConfig({
{
src: '/assets/icon-192x192.png',
sizes: '192x192',
- type: 'image/png'
+ type: 'image/png',
},
{
src: '/assets/maskable-icon.png',
@@ -113,10 +112,7 @@ export default defineConfig({
if (id.includes('node_modules/highlight.js')) {
return 'markdown_highlight';
}
- if (
- id.includes('node_modules/hast-util-raw') ||
- id.includes('node_modules/katex')
- ) {
+ if (id.includes('node_modules/hast-util-raw') || id.includes('node_modules/katex')) {
return 'markdown_large';
}
// Group TanStack libraries together.
@@ -141,10 +137,7 @@ export default defineConfig({
entryFileNames: 'assets/[name].[hash].js',
chunkFileNames: 'assets/[name].[hash].js',
assetFileNames: (assetInfo) => {
- if (
- assetInfo.names &&
- /\.(woff|woff2|eot|ttf|otf)$/.test(assetInfo.names)
- ) {
+ if (assetInfo.names && /\.(woff|woff2|eot|ttf|otf)$/.test(assetInfo.names)) {
return 'assets/fonts/[name][extname]';
}
return 'assets/[name].[hash][extname]';
@@ -187,4 +180,4 @@ export function sourcemapExclude(opts?: SourcemapExclude): Plugin {
}
},
};
-}
\ No newline at end of file
+}
diff --git a/eslint.config.mjs b/eslint.config.mjs
index 4ebe862c6f..dc2567e9b9 100644
--- a/eslint.config.mjs
+++ b/eslint.config.mjs
@@ -25,6 +25,7 @@ const compat = new FlatCompat({
export default [
{
ignores: [
+ 'client/vite.config.ts',
'client/dist/**/*',
'client/public/**/*',
'client/coverage/**/*',
diff --git a/librechat.example.yaml b/librechat.example.yaml
index 637e7a5219..7913c50455 100644
--- a/librechat.example.yaml
+++ b/librechat.example.yaml
@@ -224,7 +224,6 @@ endpoints:
- name: 'OpenRouter'
# For `apiKey` and `baseURL`, you can use environment variables that you define.
# recommended environment variables:
- # Known issue: you should not use `OPENROUTER_API_KEY` as it will then override the `openAI` endpoint to use OpenRouter as well.
apiKey: '${OPENROUTER_KEY}'
baseURL: 'https://openrouter.ai/api/v1'
models:
diff --git a/package-lock.json b/package-lock.json
index a5f18acdde..c513481bf4 100644
--- a/package-lock.json
+++ b/package-lock.json
@@ -41798,7 +41798,7 @@
},
"packages/data-provider": {
"name": "librechat-data-provider",
- "version": "0.7.6992",
+ "version": "0.7.6993",
"license": "ISC",
"dependencies": {
"axios": "^1.7.7",
diff --git a/packages/data-provider/package.json b/packages/data-provider/package.json
index 9f6d2edbf2..da2859e9c8 100644
--- a/packages/data-provider/package.json
+++ b/packages/data-provider/package.json
@@ -1,6 +1,6 @@
{
"name": "librechat-data-provider",
- "version": "0.7.6992",
+ "version": "0.7.6993",
"description": "data services for librechat apps",
"main": "dist/index.js",
"module": "dist/index.es.js",
diff --git a/packages/data-provider/src/config.ts b/packages/data-provider/src/config.ts
index 56a560f2ee..d5923645e0 100644
--- a/packages/data-provider/src/config.ts
+++ b/packages/data-provider/src/config.ts
@@ -2,8 +2,8 @@ import { z } from 'zod';
import type { ZodError } from 'zod';
import type { TModelsConfig } from './types';
import { EModelEndpoint, eModelEndpointSchema } from './schemas';
-import { fileConfigSchema } from './file-config';
import { specsConfigSchema, TSpecsConfig } from './models';
+import { fileConfigSchema } from './file-config';
import { FileSources } from './types/files';
import { MCPServersSchema } from './mcp';
@@ -31,6 +31,27 @@ export const defaultRetrievalModels = [
'gpt-4-1106',
];
+export const excludedKeys = new Set([
+ 'conversationId',
+ 'title',
+ 'iconURL',
+ 'greeting',
+ 'endpoint',
+ 'endpointType',
+ 'createdAt',
+ 'updatedAt',
+ 'expiredAt',
+ 'messages',
+ 'isArchived',
+ 'tags',
+ 'user',
+ '__v',
+ '_id',
+ 'tools',
+ 'model',
+ 'files',
+]);
+
export enum SettingsViews {
default = 'default',
advanced = 'advanced',
diff --git a/packages/data-provider/src/schemas.ts b/packages/data-provider/src/schemas.ts
index f99bcf1c04..cee0230386 100644
--- a/packages/data-provider/src/schemas.ts
+++ b/packages/data-provider/src/schemas.ts
@@ -252,7 +252,8 @@ export const googleSettings = {
},
};
-const ANTHROPIC_MAX_OUTPUT = 8192;
+const ANTHROPIC_MAX_OUTPUT = 128000;
+const DEFAULT_MAX_OUTPUT = 8192;
const LEGACY_ANTHROPIC_MAX_OUTPUT = 4096;
export const anthropicSettings = {
model: {
@@ -280,16 +281,19 @@ export const anthropicSettings = {
min: 1,
max: ANTHROPIC_MAX_OUTPUT,
step: 1,
- default: ANTHROPIC_MAX_OUTPUT,
+ default: DEFAULT_MAX_OUTPUT,
reset: (modelName: string) => {
- if (modelName.includes('claude-3-5-sonnet') || modelName.includes('claude-3-7-sonnet')) {
- return ANTHROPIC_MAX_OUTPUT;
+ if (/claude-3[-.]5-sonnet/.test(modelName) || /claude-3[-.]7/.test(modelName)) {
+ return DEFAULT_MAX_OUTPUT;
}
return 4096;
},
set: (value: number, modelName: string) => {
- if (!modelName.includes('claude-3-5-sonnet') && value > LEGACY_ANTHROPIC_MAX_OUTPUT) {
+ if (
+ !(/claude-3[-.]5-sonnet/.test(modelName) || /claude-3[-.]7/.test(modelName)) &&
+ value > LEGACY_ANTHROPIC_MAX_OUTPUT
+ ) {
return LEGACY_ANTHROPIC_MAX_OUTPUT;
}
@@ -760,37 +764,8 @@ export const googleSchema = tConversationSchema
spec: true,
maxContextTokens: true,
})
- .transform((obj) => {
- return {
- ...obj,
- model: obj.model ?? google.model.default,
- modelLabel: obj.modelLabel ?? null,
- promptPrefix: obj.promptPrefix ?? null,
- examples: obj.examples ?? [{ input: { content: '' }, output: { content: '' } }],
- temperature: obj.temperature ?? google.temperature.default,
- maxOutputTokens: obj.maxOutputTokens ?? google.maxOutputTokens.default,
- topP: obj.topP ?? google.topP.default,
- topK: obj.topK ?? google.topK.default,
- iconURL: obj.iconURL ?? undefined,
- greeting: obj.greeting ?? undefined,
- spec: obj.spec ?? undefined,
- maxContextTokens: obj.maxContextTokens ?? undefined,
- };
- })
- .catch(() => ({
- model: google.model.default,
- modelLabel: null,
- promptPrefix: null,
- examples: [{ input: { content: '' }, output: { content: '' } }],
- temperature: google.temperature.default,
- maxOutputTokens: google.maxOutputTokens.default,
- topP: google.topP.default,
- topK: google.topK.default,
- iconURL: undefined,
- greeting: undefined,
- spec: undefined,
- maxContextTokens: undefined,
- }));
+ .transform((obj: Partial<TConversation>) => removeNullishValues(obj))
+ .catch(() => ({}));
/**
* TODO: Map the following fields: