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

🔧 fix: Update Token Calculations/Mapping, MCP `env` Initialization (#6406)

Danny Avilaefb616d60016
diff --git a/api/app/clients/BaseClient.js b/api/app/clients/BaseClient.js
index e24bffa18b..d3077e68f5 100644
--- a/api/app/clients/BaseClient.js
+++ b/api/app/clients/BaseClient.js
@@ -366,17 +366,14 @@ class BaseClient {
* context: TMessage[],
* remainingContextTokens: number,
* messagesToRefine: TMessage[],
- * summaryIndex: number,
- * }>} An object with four properties: `context`, `summaryIndex`, `remainingContextTokens`, and `messagesToRefine`.
+ * }>} An object with three properties: `context`, `remainingContextTokens`, and `messagesToRefine`.
* `context` is an array of messages that fit within the token limit.
- * `summaryIndex` is the index of the first message in the `messagesToRefine` array.
* `remainingContextTokens` is the number of tokens remaining within the limit after adding the messages to the context.
* `messagesToRefine` is an array of messages that were not added to the context because they would have exceeded the token limit.
*/
async getMessagesWithinTokenLimit({ messages: _messages, maxContextTokens, instructions }) {
// Every reply is primed with <|start|>assistant<|message|>, so we
// start with 3 tokens for the label after all messages have been counted.
- let summaryIndex = -1;
let currentTokenCount = 3;
const instructionsTokenCount = instructions?.tokenCount ?? 0;
let remainingContextTokens =
@@ -409,14 +406,12 @@ class BaseClient {
}
const prunedMemory = messages;
- summaryIndex = prunedMemory.length - 1;
remainingContextTokens -= currentTokenCount;
return {
context: context.reverse(),
remainingContextTokens,
messagesToRefine: prunedMemory,
- summaryIndex,
};
}
@@ -459,7 +454,7 @@ class BaseClient {
let orderedWithInstructions = this.addInstructions(orderedMessages, instructions);
- let { context, remainingContextTokens, messagesToRefine, summaryIndex } =
+ let { context, remainingContextTokens, messagesToRefine } =
await this.getMessagesWithinTokenLimit({
messages: orderedWithInstructions,
instructions,
@@ -529,7 +524,7 @@ class BaseClient {
}
// Make sure to only continue summarization logic if the summary message was generated
- shouldSummarize = summaryMessage && shouldSummarize;
+ shouldSummarize = summaryMessage != null && shouldSummarize === true;
logger.debug('[BaseClient] Context Count (2/2)', {
remainingContextTokens,
@@ -539,17 +534,18 @@ class BaseClient {
/** @type {Record<string, number> | undefined} */
let tokenCountMap;
if (buildTokenMap) {
- tokenCountMap = orderedWithInstructions.reduce((map, message, index) => {
+ const currentPayload = shouldSummarize ? orderedWithInstructions : context;
+ tokenCountMap = currentPayload.reduce((map, message, index) => {
const { messageId } = message;
if (!messageId) {
return map;
}
- if (shouldSummarize && index === summaryIndex && !usePrevSummary) {
+ if (shouldSummarize && index === messagesToRefine.length - 1 && !usePrevSummary) {
map.summaryMessage = { ...summaryMessage, messageId, tokenCount: summaryTokenCount };
}
- map[messageId] = orderedWithInstructions[index].tokenCount;
+ map[messageId] = currentPayload[index].tokenCount;
return map;
}, {});
}
diff --git a/api/app/clients/specs/BaseClient.test.js b/api/app/clients/specs/BaseClient.test.js
index 0dae5b14d3..c9be50d3de 100644
--- a/api/app/clients/specs/BaseClient.test.js
+++ b/api/app/clients/specs/BaseClient.test.js
@@ -164,7 +164,7 @@ describe('BaseClient', () => {
const result = await TestClient.getMessagesWithinTokenLimit({ messages });
expect(result.context).toEqual(expectedContext);
- expect(result.summaryIndex).toEqual(expectedIndex);
+ expect(result.messagesToRefine.length - 1).toEqual(expectedIndex);
expect(result.remainingContextTokens).toBe(expectedRemainingContextTokens);
expect(result.messagesToRefine).toEqual(expectedMessagesToRefine);
});
@@ -200,7 +200,7 @@ describe('BaseClient', () => {
const result = await TestClient.getMessagesWithinTokenLimit({ messages });
expect(result.context).toEqual(expectedContext);
- expect(result.summaryIndex).toEqual(expectedIndex);
+ expect(result.messagesToRefine.length - 1).toEqual(expectedIndex);
expect(result.remainingContextTokens).toBe(expectedRemainingContextTokens);
expect(result.messagesToRefine).toEqual(expectedMessagesToRefine);
});
diff --git a/api/server/services/AppService.js b/api/server/services/AppService.js
index 1accd7eba6..52c646ec8a 100644
--- a/api/server/services/AppService.js
+++ b/api/server/services/AppService.js
@@ -2,6 +2,7 @@ const {
FileSources,
EModelEndpoint,
loadOCRConfig,
+ processMCPEnv,
getConfigDefaults,
} = require('librechat-data-provider');
const { checkVariables, checkHealth, checkConfig, checkAzureVariables } = require('./start/checks');
@@ -54,7 +55,7 @@ const AppService = async (app) => {
if (config.mcpServers != null) {
const mcpManager = await getMCPManager();
- await mcpManager.initializeMCP(config.mcpServers);
+ await mcpManager.initializeMCP(config.mcpServers, processMCPEnv);
await mcpManager.mapAvailableTools(availableTools);
}
diff --git a/api/server/services/Endpoints/agents/initialize.js b/api/server/services/Endpoints/agents/initialize.js
index cb539d3bc7..dec69d862c 100644
--- a/api/server/services/Endpoints/agents/initialize.js
+++ b/api/server/services/Endpoints/agents/initialize.js
@@ -178,6 +178,7 @@ const initializeAgentOptions = async ({
agent.provider = options.provider;
}
+ /** @type {import('@librechat/agents').ClientOptions} */
agent.model_parameters = Object.assign(model_parameters, options.llmConfig);
if (options.configOptions) {
agent.model_parameters.configuration = options.configOptions;
@@ -196,6 +197,7 @@ const initializeAgentOptions = async ({
const tokensModel =
agent.provider === EModelEndpoint.azureOpenAI ? agent.model : agent.model_parameters.model;
+ const maxTokens = agent.model_parameters.maxOutputTokens ?? agent.model_parameters.maxTokens ?? 0;
return {
...agent,
@@ -204,7 +206,7 @@ const initializeAgentOptions = async ({
toolContextMap,
maxContextTokens:
agent.max_context_tokens ??
- (getModelMaxTokens(tokensModel, providerEndpointMap[provider]) ?? 4000) * 0.9,
+ ((getModelMaxTokens(tokensModel, providerEndpointMap[provider]) ?? 4000) - maxTokens) * 0.9,
};
};
diff --git a/package-lock.json b/package-lock.json
index cdd6589237..b04385f985 100644
--- a/package-lock.json
+++ b/package-lock.json
@@ -40808,7 +40808,7 @@
},
"packages/data-provider": {
"name": "librechat-data-provider",
- "version": "0.7.72",
+ "version": "0.7.73",
"license": "ISC",
"dependencies": {
"axios": "^1.8.2",
diff --git a/packages/data-provider/package.json b/packages/data-provider/package.json
index a90c3afdb7..6f495a9aeb 100644
--- a/packages/data-provider/package.json
+++ b/packages/data-provider/package.json
@@ -1,6 +1,6 @@
{
"name": "librechat-data-provider",
- "version": "0.7.72",
+ "version": "0.7.73",
"description": "data services for librechat apps",
"main": "dist/index.js",
"module": "dist/index.es.js",
diff --git a/packages/data-provider/src/mcp.ts b/packages/data-provider/src/mcp.ts
index f0520dbec7..4f3225838f 100644
--- a/packages/data-provider/src/mcp.ts
+++ b/packages/data-provider/src/mcp.ts
@@ -85,3 +85,26 @@ export const MCPOptionsSchema = z.union([
]);
export const MCPServersSchema = z.record(z.string(), MCPOptionsSchema);
+
+export type MCPOptions = z.infer<typeof MCPOptionsSchema>;
+
+/**
+ * Recursively processes an object to replace environment variables in string values
+ * @param {MCPOptions} obj - The object to process
+ * @returns {MCPOptions} - The processed object with environment variables replaced
+ */
+export function processMCPEnv(obj: MCPOptions): MCPOptions {
+ if (obj === null || obj === undefined) {
+ return obj;
+ }
+
+ if ('env' in obj && obj.env) {
+ const processedEnv: Record<string, string> = {};
+ for (const [key, value] of Object.entries(obj.env)) {
+ processedEnv[key] = extractEnvVariable(value);
+ }
+ obj.env = processedEnv;
+ }
+
+ return obj;
+}
diff --git a/packages/mcp/src/manager.ts b/packages/mcp/src/manager.ts
index 3e216b7336..9fccd9de08 100644
--- a/packages/mcp/src/manager.ts
+++ b/packages/mcp/src/manager.ts
@@ -1,5 +1,5 @@
import { CallToolResultSchema } from '@modelcontextprotocol/sdk/types.js';
-import type { JsonSchemaType } from 'librechat-data-provider';
+import type { JsonSchemaType, MCPOptions } from 'librechat-data-provider';
import type { Logger } from 'winston';
import type * as t from './types/mcp';
import { formatToolContent } from './parsers';
@@ -31,13 +31,17 @@ export class MCPManager {
return MCPManager.instance;
}
- public async initializeMCP(mcpServers: t.MCPServers): Promise<void> {
+ public async initializeMCP(
+ mcpServers: t.MCPServers,
+ processMCPEnv?: (obj: MCPOptions) => MCPOptions,
+ ): Promise<void> {
this.logger.info('[MCP] Initializing servers');
const entries = Object.entries(mcpServers);
const initializedServers = new Set();
const connectionResults = await Promise.allSettled(
- entries.map(async ([serverName, config], i) => {
+ entries.map(async ([serverName, _config], i) => {
+ const config = processMCPEnv ? processMCPEnv(_config) : _config;
const connection = new MCPConnection(serverName, config, this.logger);
connection.on('connectionChange', (state) => {