LibreChat/api/server/middleware/abortMiddleware.js
Danny Avila 37964975c1
🤖 refactor: Improve Agents Memory Usage, Bump Keyv, Grok 3 (#6850)
* chore: remove unused redis file

* chore: bump keyv dependencies, and update related imports

* refactor: Implement IoRedis client for rate limiting across middleware, as node-redis via keyv not compatible

* fix: Set max listeners to expected amount

* WIP: memory improvements

* refactor: Simplify getAbortData assignment in createAbortController

* refactor: Update getAbortData to use WeakRef for content management

* WIP: memory improvements in agent chat requests

* refactor: Enhance memory management with finalization registry and cleanup functions

* refactor: Simplify domainParser calls by removing unnecessary request parameter

* refactor: Update parameter types for action tools and agent loading functions to use minimal configs

* refactor: Simplify domainParser tests by removing unnecessary request parameter

* refactor: Simplify domainParser call by removing unnecessary request parameter

* refactor: Enhance client disposal by nullifying additional properties to improve memory management

* refactor: Improve title generation by adding abort controller and timeout handling, consolidate request cleanup

* refactor: Update checkIdleConnections to skip current user when checking for idle connections if passed

* refactor: Update createMCPTool to derive userId from config and handle abort signals

* refactor: Introduce createTokenCounter function and update tokenCounter usage; enhance disposeClient to reset Graph values

* refactor: Update getMCPManager to accept userId parameter for improved idle connection handling

* refactor: Extract logToolError function for improved error handling in AgentClient

* refactor: Update disposeClient to clear handlerRegistry and graphRunnable references in client.run

* refactor: Extract createHandleNewToken function to streamline token handling in initializeClient

* chore: bump @librechat/agents

* refactor: Improve timeout handling in addTitle function for better error management

* refactor: Introduce createFetch instead of using class method

* refactor: Enhance client disposal and request data handling in AskController and EditController

* refactor: Update import statements for AnthropicClient and OpenAIClient to use specific paths

* refactor: Use WeakRef for response handling in SplitStreamHandler to prevent memory leaks

* refactor: Simplify client disposal and rename getReqData to processReqData in AskController and EditController

* refactor: Improve logging structure and parameter handling in OpenAIClient

* refactor: Remove unused GraphEvents and improve stream event handling in AnthropicClient and OpenAIClient

* refactor: Simplify client initialization in AskController and EditController

* refactor: Remove unused mock functions and implement in-memory store for KeyvMongo

* chore: Update dependencies in package-lock.json to latest versions

* refactor: Await token usage recording in OpenAIClient to ensure proper async handling

* refactor: Remove handleAbort route from multiple endpoints and enhance client disposal logic

* refactor: Enhance abort controller logic by managing abortKey more effectively

* refactor: Add newConversation handling in useEventHandlers for improved conversation management

* fix: dropparams

* refactor: Use optional chaining for safer access to request properties in BaseClient

* refactor: Move client disposal and request data processing logic to cleanup module for better organization

* refactor: Remove aborted request check from addTitle function for cleaner logic

* feat: Add Grok 3 model pricing and update tests for new models

* chore: Remove trace warnings and inspect flags from backend start script used for debugging

* refactor: Replace user identifier handling with userId for consistency across controllers, use UserId in clientRegistry

* refactor: Enhance client disposal logic to prevent memory leaks by clearing additional references

* chore: Update @librechat/agents to version 2.4.14 in package.json and package-lock.json
2025-04-12 18:46:36 -04:00

396 lines
12 KiB
JavaScript

// abortMiddleware.js
const { isAssistantsEndpoint, ErrorTypes } = require('librechat-data-provider');
const { sendMessage, sendError, countTokens, isEnabled } = require('~/server/utils');
const { truncateText, smartTruncateText } = require('~/app/clients/prompts');
const clearPendingReq = require('~/cache/clearPendingReq');
const { spendTokens } = require('~/models/spendTokens');
const abortControllers = require('./abortControllers');
const { saveMessage, getConvo } = require('~/models');
const { abortRun } = require('./abortRun');
const { logger } = require('~/config');
const abortDataMap = new WeakMap();
function cleanupAbortController(abortKey) {
if (!abortControllers.has(abortKey)) {
return false;
}
const { abortController } = abortControllers.get(abortKey);
if (!abortController) {
abortControllers.delete(abortKey);
return true;
}
// 1. Check if this controller has any composed signals and clean them up
try {
// This creates a temporary composed signal to use for cleanup
const composedSignal = AbortSignal.any([abortController.signal]);
// Get all event types - in practice, AbortSignal typically only uses 'abort'
const eventTypes = ['abort'];
// First, execute a dummy listener removal to handle potential composed signals
for (const eventType of eventTypes) {
const dummyHandler = () => {};
composedSignal.addEventListener(eventType, dummyHandler);
composedSignal.removeEventListener(eventType, dummyHandler);
const listeners = composedSignal.listeners?.(eventType) || [];
for (const listener of listeners) {
composedSignal.removeEventListener(eventType, listener);
}
}
} catch (e) {
logger.debug(`Error cleaning up composed signals: ${e}`);
}
// 2. Abort the controller if not already aborted
if (!abortController.signal.aborted) {
abortController.abort();
}
// 3. Remove from registry
abortControllers.delete(abortKey);
// 4. Clean up any data stored in the WeakMap
if (abortDataMap.has(abortController)) {
abortDataMap.delete(abortController);
}
// 5. Clean up function references on the controller
if (abortController.getAbortData) {
abortController.getAbortData = null;
}
if (abortController.abortCompletion) {
abortController.abortCompletion = null;
}
return true;
}
async function abortMessage(req, res) {
let { abortKey, endpoint } = req.body;
if (isAssistantsEndpoint(endpoint)) {
return await abortRun(req, res);
}
const conversationId = abortKey?.split(':')?.[0] ?? req.user.id;
if (!abortControllers.has(abortKey) && abortControllers.has(conversationId)) {
abortKey = conversationId;
}
if (!abortControllers.has(abortKey) && !res.headersSent) {
return res.status(204).send({ message: 'Request not found' });
}
const { abortController } = abortControllers.get(abortKey) ?? {};
if (!abortController) {
return res.status(204).send({ message: 'Request not found' });
}
const finalEvent = await abortController.abortCompletion?.();
logger.debug(
`[abortMessage] ID: ${req.user.id} | ${req.user.email} | Aborted request: ` +
JSON.stringify({ abortKey }),
);
cleanupAbortController(abortKey);
if (res.headersSent && finalEvent) {
return sendMessage(res, finalEvent);
}
res.setHeader('Content-Type', 'application/json');
res.send(JSON.stringify(finalEvent));
}
const handleAbort = () => {
return async (req, res) => {
try {
if (isEnabled(process.env.LIMIT_CONCURRENT_MESSAGES)) {
await clearPendingReq({ userId: req.user.id });
}
return await abortMessage(req, res);
} catch (err) {
logger.error('[abortMessage] handleAbort error', err);
}
};
};
const createAbortController = (req, res, getAbortData, getReqData) => {
const abortController = new AbortController();
const { endpointOption } = req.body;
// Store minimal data in WeakMap to avoid circular references
abortDataMap.set(abortController, {
getAbortDataFn: getAbortData,
userId: req.user.id,
endpoint: endpointOption.endpoint,
iconURL: endpointOption.iconURL,
model: endpointOption.modelOptions?.model || endpointOption.model_parameters?.model,
});
// Replace the direct function reference with a wrapper that uses WeakMap
abortController.getAbortData = function () {
const data = abortDataMap.get(this);
if (!data || typeof data.getAbortDataFn !== 'function') {
return {};
}
try {
const result = data.getAbortDataFn();
// Create a copy without circular references
const cleanResult = { ...result };
// If userMessagePromise exists, break its reference to client
if (
cleanResult.userMessagePromise &&
typeof cleanResult.userMessagePromise.then === 'function'
) {
// Create a new promise that fulfills with the same result but doesn't reference the original
const originalPromise = cleanResult.userMessagePromise;
cleanResult.userMessagePromise = new Promise((resolve, reject) => {
originalPromise.then(
(result) => resolve({ ...result }),
(error) => reject(error),
);
});
}
return cleanResult;
} catch (err) {
logger.error('[abortController.getAbortData] Error:', err);
return {};
}
};
/**
* @param {TMessage} userMessage
* @param {string} responseMessageId
*/
const onStart = (userMessage, responseMessageId) => {
sendMessage(res, { message: userMessage, created: true });
const abortKey = userMessage?.conversationId ?? req.user.id;
getReqData({ abortKey });
const prevRequest = abortControllers.get(abortKey);
const { overrideUserMessageId } = req?.body ?? {};
if (overrideUserMessageId != null && prevRequest && prevRequest?.abortController) {
const data = prevRequest.abortController.getAbortData();
getReqData({ userMessage: data?.userMessage });
const addedAbortKey = `${abortKey}:${responseMessageId}`;
// Store minimal options
const minimalOptions = {
endpoint: endpointOption.endpoint,
iconURL: endpointOption.iconURL,
model: endpointOption.modelOptions?.model || endpointOption.model_parameters?.model,
};
abortControllers.set(addedAbortKey, { abortController, ...minimalOptions });
// Use a simple function for cleanup to avoid capturing context
const cleanupHandler = () => {
try {
cleanupAbortController(addedAbortKey);
} catch (e) {
// Ignore cleanup errors
}
};
res.on('finish', cleanupHandler);
return;
}
// Store minimal options
const minimalOptions = {
endpoint: endpointOption.endpoint,
iconURL: endpointOption.iconURL,
model: endpointOption.modelOptions?.model || endpointOption.model_parameters?.model,
};
abortControllers.set(abortKey, { abortController, ...minimalOptions });
// Use a simple function for cleanup to avoid capturing context
const cleanupHandler = () => {
try {
cleanupAbortController(abortKey);
} catch (e) {
// Ignore cleanup errors
}
};
res.on('finish', cleanupHandler);
};
// Define abortCompletion without capturing the entire parent scope
abortController.abortCompletion = async function () {
this.abort();
// Get data from WeakMap
const ctrlData = abortDataMap.get(this);
if (!ctrlData || !ctrlData.getAbortDataFn) {
return { final: true, conversation: {}, title: 'New Chat' };
}
// Get abort data using stored function
const { conversationId, userMessage, userMessagePromise, promptTokens, ...responseData } =
ctrlData.getAbortDataFn();
const completionTokens = await countTokens(responseData?.text ?? '');
const user = ctrlData.userId;
const responseMessage = {
...responseData,
conversationId,
finish_reason: 'incomplete',
endpoint: ctrlData.endpoint,
iconURL: ctrlData.iconURL,
model: ctrlData.modelOptions?.model ?? ctrlData.model_parameters?.model,
unfinished: false,
error: false,
isCreatedByUser: false,
tokenCount: completionTokens,
};
await spendTokens(
{ ...responseMessage, context: 'incomplete', user },
{ promptTokens, completionTokens },
);
await saveMessage(
req,
{ ...responseMessage, user },
{ context: 'api/server/middleware/abortMiddleware.js' },
);
let conversation;
if (userMessagePromise) {
const resolved = await userMessagePromise;
conversation = resolved?.conversation;
// Break reference to promise
resolved.conversation = null;
}
if (!conversation) {
conversation = await getConvo(user, conversationId);
}
return {
title: conversation && !conversation.title ? null : conversation?.title || 'New Chat',
final: true,
conversation,
requestMessage: userMessage,
responseMessage: responseMessage,
};
};
return { abortController, onStart };
};
/**
* @param {ServerResponse} res
* @param {ServerRequest} req
* @param {Error | unknown} error
* @param {Partial<TMessage> & { partialText?: string }} data
* @returns { Promise<void> }
*/
const handleAbortError = async (res, req, error, data) => {
if (error?.message?.includes('base64')) {
logger.error('[handleAbortError] Error in base64 encoding', {
...error,
stack: smartTruncateText(error?.stack, 1000),
message: truncateText(error.message, 350),
});
} else {
logger.error('[handleAbortError] AI response error; aborting request:', error);
}
const { sender, conversationId, messageId, parentMessageId, partialText } = data;
if (error.stack && error.stack.includes('google')) {
logger.warn(
`AI Response error for conversation ${conversationId} likely caused by Google censor/filter`,
);
}
let errorText = error?.message?.includes('"type"')
? error.message
: 'An error occurred while processing your request. Please contact the Admin.';
if (error?.type === ErrorTypes.INVALID_REQUEST) {
errorText = `{"type":"${ErrorTypes.INVALID_REQUEST}"}`;
}
if (error?.message?.includes('does not support \'system\'')) {
errorText = `{"type":"${ErrorTypes.NO_SYSTEM_MESSAGES}"}`;
}
/**
* @param {string} partialText
* @returns {Promise<void>}
*/
const respondWithError = async (partialText) => {
const endpointOption = req.body?.endpointOption;
let options = {
sender,
messageId,
conversationId,
parentMessageId,
text: errorText,
user: req.user.id,
shouldSaveMessage: true,
spec: endpointOption?.spec,
iconURL: endpointOption?.iconURL,
modelLabel: endpointOption?.modelLabel,
model: endpointOption?.modelOptions?.model || req.body?.model,
};
if (req.body?.agent_id) {
options.agent_id = req.body.agent_id;
}
if (partialText) {
options = {
...options,
error: false,
unfinished: true,
text: partialText,
};
}
// Create a simple callback without capturing parent scope
const callback = async () => {
try {
cleanupAbortController(conversationId);
} catch (e) {
// Ignore cleanup errors
}
};
await sendError(req, res, options, callback);
};
if (partialText && partialText.length > 5) {
try {
return await abortMessage(req, res);
} catch (err) {
logger.error('[handleAbortError] error while trying to abort message', err);
return respondWithError(partialText);
}
} else {
return respondWithError();
}
};
module.exports = {
handleAbort,
handleAbortError,
createAbortController,
cleanupAbortController,
};