Merge branch 'main' into main

api/app/clients/BaseClient.js

@@ -649,15 +649,17 @@ class BaseClient {
 
     this.responsePromise = this.saveMessageToDatabase(responseMessage, saveOptions, user);
     this.savedMessageIds.add(responseMessage.messageId);
-    const messageCache = getLogStores(CacheKeys.MESSAGES);
-    messageCache.set(
-      responseMessageId,
-      {
-        text: responseMessage.text,
-        complete: true,
-      },
-      Time.FIVE_MINUTES,
-    );
+    if (responseMessage.text) {
+      const messageCache = getLogStores(CacheKeys.MESSAGES);
+      messageCache.set(
+        responseMessageId,
+        {
+          text: responseMessage.text,
+          complete: true,
+        },
+        Time.FIVE_MINUTES,
+      );
+    }
     delete responseMessage.tokenCount;
     return responseMessage;
   }

api/app/clients/PluginsClient.js

@@ -256,15 +256,17 @@ class PluginsClient extends OpenAIClient {
     }
 
     this.responsePromise = this.saveMessageToDatabase(responseMessage, saveOptions, user);
-    const messageCache = getLogStores(CacheKeys.MESSAGES);
-    messageCache.set(
-      responseMessage.messageId,
-      {
-        text: responseMessage.text,
-        complete: true,
-      },
-      Time.FIVE_MINUTES,
-    );
+    if (responseMessage.text) {
+      const messageCache = getLogStores(CacheKeys.MESSAGES);
+      messageCache.set(
+        responseMessage.messageId,
+        {
+          text: responseMessage.text,
+          complete: true,
+        },
+        Time.FIVE_MINUTES,
+      );
+    }
     delete responseMessage.tokenCount;
     return { ...responseMessage, ...result };
   }

api/cache/getLogStores.js

@@ -5,49 +5,51 @@ const { math, isEnabled } = require('~/server/utils');
 const keyvRedis = require('./keyvRedis');
 const keyvMongo = require('./keyvMongo');
 
-const { BAN_DURATION, USE_REDIS } = process.env ?? {};
+const { BAN_DURATION, USE_REDIS, DEBUG_MEMORY_CACHE } = process.env ?? {};
 
 const duration = math(BAN_DURATION, 7200000);
+const isRedisEnabled = isEnabled(USE_REDIS);
+const debugMemoryCache = isEnabled(DEBUG_MEMORY_CACHE);
 
 const createViolationInstance = (namespace) => {
-  const config = isEnabled(USE_REDIS) ? { store: keyvRedis } : { store: violationFile, namespace };
+  const config = isRedisEnabled ? { store: keyvRedis } : { store: violationFile, namespace };
   return new Keyv(config);
 };
 
 // Serve cache from memory so no need to clear it on startup/exit
-const pending_req = isEnabled(USE_REDIS)
+const pending_req = isRedisEnabled
   ? new Keyv({ store: keyvRedis })
   : new Keyv({ namespace: 'pending_req' });
 
-const config = isEnabled(USE_REDIS)
+const config = isRedisEnabled
   ? new Keyv({ store: keyvRedis })
   : new Keyv({ namespace: CacheKeys.CONFIG_STORE });
 
-const roles = isEnabled(USE_REDIS)
+const roles = isRedisEnabled
   ? new Keyv({ store: keyvRedis })
   : new Keyv({ namespace: CacheKeys.ROLES });
 
-const audioRuns = isEnabled(USE_REDIS)
+const audioRuns = isRedisEnabled
   ? new Keyv({ store: keyvRedis, ttl: Time.TEN_MINUTES })
   : new Keyv({ namespace: CacheKeys.AUDIO_RUNS, ttl: Time.TEN_MINUTES });
 
-const messages = isEnabled(USE_REDIS)
-  ? new Keyv({ store: keyvRedis, ttl: Time.FIVE_MINUTES })
-  : new Keyv({ namespace: CacheKeys.MESSAGES, ttl: Time.FIVE_MINUTES });
+const messages = isRedisEnabled
+  ? new Keyv({ store: keyvRedis, ttl: Time.ONE_MINUTE })
+  : new Keyv({ namespace: CacheKeys.MESSAGES, ttl: Time.ONE_MINUTE });
 
-const tokenConfig = isEnabled(USE_REDIS)
+const tokenConfig = isRedisEnabled
   ? new Keyv({ store: keyvRedis, ttl: Time.THIRTY_MINUTES })
   : new Keyv({ namespace: CacheKeys.TOKEN_CONFIG, ttl: Time.THIRTY_MINUTES });
 
-const genTitle = isEnabled(USE_REDIS)
+const genTitle = isRedisEnabled
   ? new Keyv({ store: keyvRedis, ttl: Time.TWO_MINUTES })
   : new Keyv({ namespace: CacheKeys.GEN_TITLE, ttl: Time.TWO_MINUTES });
 
 const modelQueries = isEnabled(process.env.USE_REDIS)
   ? new Keyv({ store: keyvRedis })
   : new Keyv({ namespace: CacheKeys.MODEL_QUERIES });
 
-const abortKeys = isEnabled(USE_REDIS)
+const abortKeys = isRedisEnabled
   ? new Keyv({ store: keyvRedis })
   : new Keyv({ namespace: CacheKeys.ABORT_KEYS, ttl: Time.TEN_MINUTES });
 
@@ -88,6 +90,154 @@ const namespaces = {
   [CacheKeys.MESSAGES]: messages,
 };
 
+/**
+ * Gets all cache stores that have TTL configured
+ * @returns {Keyv[]}
+ */
+function getTTLStores() {
+  return Object.values(namespaces).filter((store) =>
+    store instanceof Keyv &&
+    typeof store.opts?.ttl === 'number' &&
+    store.opts.ttl > 0,
+  );
+}
+
+/**
+ * Clears entries older than the cache's TTL
+ * @param {Keyv} cache
+ */
+async function clearExpiredFromCache(cache) {
+  if (!cache?.opts?.store?.entries) {
+    return;
+  }
+
+  const ttl = cache.opts.ttl;
+  if (!ttl) {
+    return;
+  }
+
+  const expiryTime = Date.now() - ttl;
+  let cleared = 0;
+
+  // Get all keys first to avoid modification during iteration
+  const keys = Array.from(cache.opts.store.keys());
+
+  for (const key of keys) {
+    try {
+      const raw = cache.opts.store.get(key);
+      if (!raw) {continue;}
+
+      const data = cache.opts.deserialize(raw);
+      // Check if the entry is older than TTL
+      if (data?.expires && data.expires <= expiryTime) {
+        const deleted = await cache.opts.store.delete(key);
+        if (!deleted) {
+          debugMemoryCache && console.warn(`[Cache] Error deleting entry: ${key} from ${cache.opts.namespace}`);
+          continue;
+        }
+        cleared++;
+      }
+    } catch (error) {
+      debugMemoryCache && console.log(`[Cache] Error processing entry from ${cache.opts.namespace}:`, error);
+      const deleted = await cache.opts.store.delete(key);
+      if (!deleted) {
+        debugMemoryCache && console.warn(`[Cache] Error deleting entry: ${key} from ${cache.opts.namespace}`);
+        continue;
+      }
+      cleared++;
+    }
+  }
+
+  if (cleared > 0) {
+    debugMemoryCache && console.log(`[Cache] Cleared ${cleared} entries older than ${ttl}ms from ${cache.opts.namespace}`);
+  }
+}
+
+const auditCache = () => {
+  const ttlStores = getTTLStores();
+  console.log('[Cache] Starting audit');
+
+  ttlStores.forEach(store => {
+    if (!store?.opts?.store?.entries) {
+      return;
+    }
+
+    console.log(`[Cache] ${store.opts.namespace} entries:`, {
+      count: store.opts.store.size,
+      ttl: store.opts.ttl,
+      keys: Array.from(store.opts.store.keys()),
+      entriesWithTimestamps: Array.from(store.opts.store.entries())
+        .map(([key, value]) => ({
+          key,
+          value,
+        })),
+    });
+  });
+};
+
+/**
+   * Clears expired entries from all TTL-enabled stores
+   */
+async function clearAllExpiredFromCache() {
+  const ttlStores = getTTLStores();
+  await Promise.all(ttlStores.map(store => clearExpiredFromCache(store)));
+
+  // Force garbage collection if available (Node.js with --expose-gc flag)
+  if (global.gc) {
+    global.gc();
+  }
+}
+
+if (!isRedisEnabled) {
+  /** @type {Set<NodeJS.Timeout>} */
+  const cleanupIntervals = new Set();
+
+  // Clear expired entries every 30 seconds
+  const cleanup = setInterval(() => {
+    clearAllExpiredFromCache();
+  }, Time.THIRTY_SECONDS);
+
+  cleanupIntervals.add(cleanup);
+
+  if (debugMemoryCache) {
+    const monitor = setInterval(() => {
+      const ttlStores = getTTLStores();
+      const memory = process.memoryUsage();
+      const totalSize = ttlStores.reduce((sum, store) => sum + (store.opts?.store?.size ?? 0), 0);
+
+      console.log('[Cache] Memory usage:', {
+        heapUsed: `${(memory.heapUsed / 1024 / 1024).toFixed(2)} MB`,
+        heapTotal: `${(memory.heapTotal / 1024 / 1024).toFixed(2)} MB`,
+        rss: `${(memory.rss / 1024 / 1024).toFixed(2)} MB`,
+        external: `${(memory.external / 1024 / 1024).toFixed(2)} MB`,
+        totalCacheEntries: totalSize,
+      });
+
+      auditCache();
+    }, Time.ONE_MINUTE);
+
+    cleanupIntervals.add(monitor);
+  }
+
+  const dispose = () => {
+    debugMemoryCache && console.log('[Cache] Cleaning up and shutting down...');
+    cleanupIntervals.forEach(interval => clearInterval(interval));
+    cleanupIntervals.clear();
+
+    // One final cleanup before exit
+    clearAllExpiredFromCache().then(() => {
+      debugMemoryCache && console.log('[Cache] Final cleanup completed');
+      process.exit(0);
+    });
+  };
+
+  // Handle various termination signals
+  process.on('SIGTERM', dispose);
+  process.on('SIGINT', dispose);
+  process.on('SIGQUIT', dispose);
+  process.on('SIGHUP', dispose);
+}
+
 /**
  * Returns the keyv cache specified by type.
  * If an invalid type is passed, an error will be thrown.

api/models/Transaction.js

@@ -27,6 +27,9 @@ transactionSchema.methods.calculateTokenValue = function () {
  */
 transactionSchema.statics.create = async function (txData) {
   const Transaction = this;
+  if (txData.rawAmount != null && isNaN(txData.rawAmount)) {
+    return;
+  }
 
   const transaction = new Transaction(txData);
   transaction.endpointTokenConfig = txData.endpointTokenConfig;

api/models/Transaction.spec.js

@@ -1,5 +1,6 @@
 const mongoose = require('mongoose');
 const { MongoMemoryServer } = require('mongodb-memory-server');
+const { Transaction } = require('./Transaction');
 const Balance = require('./Balance');
 const { spendTokens, spendStructuredTokens } = require('./spendTokens');
 const { getMultiplier, getCacheMultiplier } = require('./tx');
@@ -346,3 +347,28 @@ describe('Structured Token Spending Tests', () => {
     expect(result.completion.completion).toBeCloseTo(-50 * 15 * 1.15, 0); // Assuming multiplier is 15 and cancelRate is 1.15
   });
 });
+
+describe('NaN Handling Tests', () => {
+  test('should skip transaction creation when rawAmount is NaN', async () => {
+    const userId = new mongoose.Types.ObjectId();
+    const initialBalance = 10000000;
+    await Balance.create({ user: userId, tokenCredits: initialBalance });
+
+    const model = 'gpt-3.5-turbo';
+    const txData = {
+      user: userId,
+      conversationId: 'test-conversation-id',
+      model,
+      context: 'test',
+      endpointTokenConfig: null,
+      rawAmount: NaN,
+      tokenType: 'prompt',
+    };
+
+    const result = await Transaction.create(txData);
+    expect(result).toBeUndefined();
+
+    const balance = await Balance.findOne({ user: userId });
+    expect(balance.tokenCredits).toBe(initialBalance);
+  });
+});

api/package.json

@@ -44,7 +44,7 @@
     "@langchain/google-genai": "^0.1.4",
     "@langchain/google-vertexai": "^0.1.4",
     "@langchain/textsplitters": "^0.1.0",
-    "@librechat/agents": "^1.8.8",
+    "@librechat/agents": "^1.9.7",
     "axios": "^1.7.7",
     "bcryptjs": "^2.4.3",
     "cheerio": "^1.0.0-rc.12",

api/server/controllers/agents/callbacks.js

@@ -1,8 +1,10 @@
 const { Tools, StepTypes, imageGenTools, FileContext } = require('librechat-data-provider');
 const {
   EnvVar,
+  Providers,
   GraphEvents,
   ToolEndHandler,
+  handleToolCalls,
   ChatModelStreamHandler,
 } = require('@librechat/agents');
 const { processCodeOutput } = require('~/server/services/Files/Code/process');
@@ -57,13 +59,22 @@ class ModelEndHandler {
       return;
     }
 
-    const usage = data?.output?.usage_metadata;
-    if (metadata?.model) {
-      usage.model = metadata.model;
-    }
+    try {
+      if (metadata.provider === Providers.GOOGLE) {
+        handleToolCalls(data?.output?.tool_calls, metadata, graph);
+      }
+
+      const usage = data?.output?.usage_metadata;
+      if (!usage) {
+        return;
+      }
+      if (metadata?.model) {
+        usage.model = metadata.model;
+      }
 
-    if (usage) {
       this.collectedUsage.push(usage);
+    } catch (error) {
+      logger.error('Error handling model end event:', error);
     }
   }
 }

api/server/controllers/assistants/chatV2.js

@@ -398,15 +398,17 @@ const chatV2 = async (req, res) => {
       response = streamRunManager;
       response.text = streamRunManager.intermediateText;
 
-      const messageCache = getLogStores(CacheKeys.MESSAGES);
-      messageCache.set(
-        responseMessageId,
-        {
-          complete: true,
-          text: response.text,
-        },
-        Time.FIVE_MINUTES,
-      );
+      if (response.text) {
+        const messageCache = getLogStores(CacheKeys.MESSAGES);
+        messageCache.set(
+          responseMessageId,
+          {
+            complete: true,
+            text: response.text,
+          },
+          Time.FIVE_MINUTES,
+        );
+      }
     };
 
     await processRun();