@peopl-health/nexus 2.5.1 → 2.5.2

package/lib/core/NexusMessaging.js

@@ -1,4 +1,5 @@
 const { airtable, getBase } = require('../config/airtableConfig');
+const { Message } = require('../models/messageModel');
 const { addMsgAssistant, replyAssistant } = require('../services/assistantService');
 const { createProvider } = require('../adapters/registry');
 const runtimeConfig = require('../config/runtimeConfig');
@@ -42,7 +43,7 @@ class NexusMessaging {
     this.pendingResponses = new Map();
     this.batchingConfig = {
       enabled: config.messageBatching?.enabled ?? false,
-      baseWaitTime: config.messageBatching?.baseWaitTime ?? 15000
+      baseWaitTime: config.messageBatching?.baseWaitTime ?? 10000
     };
   }
 
@@ -353,6 +354,18 @@ class NexusMessaging {
 
     const chatId = messageData.from || messageData.From;
 
+    // Send initial typing indicator with delay
+    if (chatId && this.provider && typeof this.provider.sendTypingIndicator === 'function') {
+      const messageId = messageData.id || messageData.MessageSid || messageData.message_id;        
+      if (messageId) {
+        setTimeout(() => {
+          this.provider.sendTypingIndicator(messageId).catch(err => 
+            logger.debug('[processIncomingMessage] Typing indicator failed', { error: err.message })
+          );
+        }, 3000);
+      }
+    }
+
     if (chatId && hasPreprocessingHandler()) {
       const stop = await invokePreprocessingHandler({
         code: chatId,
@@ -372,28 +385,11 @@ class NexusMessaging {
     } else if (messageData.flow) {
       return await this.handleFlow(messageData);
     } else {
-      if (chatId && this.provider && typeof this.provider.sendTypingIndicator === 'function') {
-        const messageId = messageData.id || messageData.MessageSid || messageData.message_id;        
-        if (messageId) {
-          this.provider.sendTypingIndicator(messageId)
-            .then(() => logger.debug('[processIncomingMessage] Typing indicator sent successfully', { messageId }))
-            .catch(err => logger.warn('[processIncomingMessage] Typing indicator failed', { 
-              error: err.message, 
-              messageId 
-            }));
-        } else {
-          logger.warn('[processIncomingMessage] No message ID found for typing indicator', { 
-            messageData: JSON.stringify(messageData).substring(0, 200) 
-          });
-        }
-      }
-      
       // For regular messages and media, use batching if enabled
       logger.info('Batching config:', this.batchingConfig);
       if (this.batchingConfig.enabled && chatId) {
         return await this._handleWithBatching(messageData, chatId);
       } else {
-        // Handle media and regular messages without batching
         if (messageData.media) {
           return await this.handleMedia(messageData);
         } else {
@@ -650,14 +646,24 @@ class NexusMessaging {
    * Handle message with batching - waits for additional messages before processing
    */
   async _handleWithBatching(messageData, chatId) {
-    if (this.pendingResponses.has(chatId)) {
-      clearTimeout(this.pendingResponses.get(chatId));
+    const existing = this.pendingResponses.get(chatId);
+    if (existing) {
+      clearTimeout(existing.timeoutId);
+      if (existing.typingInterval) {
+        clearInterval(existing.typingInterval);
+      }
       logger.info(`Received additional message from ${chatId}, resetting wait timer`);
     }
 
+    // Start typing indicator refresh for batching period
+    const typingInterval = await this._startTypingRefresh(chatId);
+
     const waitTime = this.batchingConfig.baseWaitTime;
     const timeoutId = setTimeout(async () => {
       try {
+        if (typingInterval) {
+          clearInterval(typingInterval);
+        }
         this.pendingResponses.delete(chatId);
         await this._handleBatchedMessages(chatId);
       } catch (error) {
@@ -665,17 +671,44 @@ class NexusMessaging {
       }
     }, waitTime);
     
-    this.pendingResponses.set(chatId, timeoutId);
+    this.pendingResponses.set(chatId, { timeoutId, typingInterval });
     logger.info(`Waiting ${Math.round(waitTime/1000)} seconds for more messages from ${chatId}`);
   }
 
+  /**
+   * Start typing indicator refresh interval
+   */
+  async _startTypingRefresh(chatId) {
+    if (!this.provider || typeof this.provider.sendTypingIndicator !== 'function') {
+      return null;
+    }
+
+    const lastMessage = await Message.findOne({ 
+      numero: chatId, 
+      from_me: false,
+      message_id: { $exists: true, $ne: null }
+    }).sort({ createdAt: -1 });
+    
+    if (!lastMessage?.message_id) return null;
+    
+    return setInterval(() => 
+      this.provider.sendTypingIndicator(lastMessage.message_id).catch(err => 
+        logger.debug('[_startTypingRefresh] Failed', { error: err.message })
+      ), 5000
+    );
+  }
+
   /**
    * Process all batched messages for a chat
    */
   async _handleBatchedMessages(chatId) {
+    let typingInterval = null;
+    
     try {
       logger.info(`Processing batched messages from ${chatId} (including media if any)`);
       
+      typingInterval = await this._startTypingRefresh(chatId);
+      
       // Get assistant response
       const result = await replyAssistant(chatId);
       const botResponse = typeof result === 'string' ? result : result?.output;
@@ -696,6 +729,10 @@ class NexusMessaging {
       
     } catch (error) {
       logger.error('Error in batched message handling:', { error: error.message });
+    } finally {
+      if (typingInterval) {
+        clearInterval(typingInterval);
+      }
     }
   }
 
@@ -703,8 +740,14 @@ class NexusMessaging {
    * Clear pending response for a chat (useful for cleanup)
    */
   clearPendingResponse(chatId) {
-    if (this.pendingResponses.has(chatId)) {
-      clearTimeout(this.pendingResponses.get(chatId));
+    const pending = this.pendingResponses.get(chatId);
+    if (pending) {
+      if (pending.timeoutId) {
+        clearTimeout(pending.timeoutId);
+      }
+      if (pending.typingInterval) {
+        clearInterval(pending.typingInterval);
+      }
       this.pendingResponses.delete(chatId);
     }
   }

package/lib/services/assistantService.js

@@ -1,32 +1,18 @@
-const { Historial_Clinico_ID } = require('../config/airtableConfig.js');
-const AWS = require('../config/awsConfig.js');
+const { withTracing } = require('../utils/tracingDecorator.js');
 const llmConfig = require('../config/llmConfig');
-const runtimeConfig = require('../config/runtimeConfig');
 const { BaseAssistant } = require('../assistants/BaseAssistant');
-const { createProvider } = require('../providers/createProvider');
-
-const { Message } = require('../models/messageModel.js');
-const { Thread } = require('../models/threadModel.js');
-const { PredictionMetrics } = require('../models/predictionMetricsModel');
-const { insertMessage } = require('../models/messageModel');
-
-const { getCurRow } = require('../helpers/assistantHelper.js');
-const { runAssistantAndWait, runAssistantWithRetries } = require('../helpers/assistantHelper.js');
-const { getThread, getThreadInfo } = require('../helpers/threadHelper.js');
-const { withTracing } = require('../utils/tracingDecorator.js');
-const { processThreadMessage } = require('../helpers/processHelper.js');
-const { getLastMessages, updateMessageRecord } = require('../helpers/messageHelper.js');
-const { withThreadRecovery } = require('../helpers/threadRecoveryHelper.js');
-const { combineImagesToPDF, cleanupFiles } = require('../helpers/filesHelper.js');
-const { logger } = require('../utils/logger');
-
-const DEFAULT_MAX_RETRIES = parseInt(process.env.MAX_RETRIES || '30', 10);
+const {
+  createAssistantCore,
+  addMsgAssistantCore,
+  addInstructionCore,
+  replyAssistantCore,
+  switchAssistantCore
+} = require('./assistantServiceCore');
 
 let assistantConfig = null;
 let assistantRegistry = {};
 let customGetAssistantById = null;
 
-
 const configureAssistants = (config) => {
   if (!config) {
     throw new Error('Assistant configuration is required');
@@ -99,224 +85,46 @@ const overrideGetAssistantById = (resolverFn) => {
 
 const getAssistantById = (assistant_id, thread) => {
   if (customGetAssistantById) {
-    const inst = customGetAssistantById(assistant_id, thread);
-    if (inst) return inst;
+    return customGetAssistantById(assistant_id, thread);
   }
 
-  if (!assistantConfig) {
-    assistantConfig = {};
+  if (assistantRegistry[assistant_id]) {
+    const AssistantClass = assistantRegistry[assistant_id];
+    return new AssistantClass({ thread });
   }
-  
-  const AssistantClass = assistantRegistry[assistant_id];
-  if (!AssistantClass) {
-    throw new Error(`Assistant '${assistant_id}' not found. Available assistants: ${Object.keys(assistantRegistry).join(', ')}`);
-  }
-
-  const provider = createProvider({ variant: process.env.VARIANT || 'assistants' });
-  const sharedClient = provider?.getClient?.() || llmConfig.openaiClient || null;
 
-  if (AssistantClass.prototype instanceof BaseAssistant) {
-    return new AssistantClass({
+  if (assistantConfig && assistantConfig[assistant_id]) {
+    const config = assistantConfig[assistant_id];
+    return new BaseAssistant({
+      ...config,
       assistantId: assistant_id,
-      thread,
-      client: sharedClient,
-      provider
-    });
-  }
-
-  try {
-    return new AssistantClass(thread);
-  } catch (error) {
-    return new AssistantClass({
-      thread,
-      assistantId: assistant_id,
-      client: sharedClient,
-      provider
-    });
-  }
-};
-
-
-const createAssistant = async (code, assistant_id, messages=[], force=false) => {
-  const findThread = await Thread.findOne({ code: code });
-  logger.info('[createAssistant] findThread', findThread);
-  if (findThread && findThread.getConversationId() && !force) {
-    logger.info('[createAssistant] Thread already exists');
-    const updateFields = { active: true, stopped: false };
-    Thread.setAssistantId(updateFields, assistant_id);    
-    await Thread.updateOne({ code: code }, { $set: updateFields });
-    return findThread;
-  }
-
-  if (force && findThread?.getConversationId()) {
-    const provider = createProvider({ variant: process.env.VARIANT || 'assistants' });
-    await provider.deleteConversation(findThread.getConversationId());
-    logger.info('[createAssistant] Deleted old conversation, will create new one');
-  }
-  
-  const curRow = await getCurRow(Historial_Clinico_ID, code);
-  logger.info('[createAssistant] curRow', curRow[0]);
-  const nombre = curRow?.[0]?.['name'] || null;
-  const patientId = curRow?.[0]?.['record_id'] || null;
-
-  const assistant = getAssistantById(assistant_id, null);
-  const initialThread = await assistant.create(code, curRow[0]);
-
-  const provider = createProvider({ variant: process.env.VARIANT || 'assistants' });
-  for (const message of messages) {
-    await provider.addMessage({
-      threadId: initialThread.id,
-      role: 'assistant',
-      content: message
+      thread
     });
   }
-  
-  const thread = {
-    code: code,
-    patient_id: patientId,
-    nombre: nombre,
-    active: true
-  };
-  Thread.setAssistantId(thread, assistant_id);
-  Thread.setConversationId(thread, initialThread.id);
 
-  const condition = { $or: [{ code: code }] };
-  const options = { new: true, upsert: true };
-  const updatedThread = await Thread.findOneAndUpdate(condition, {run_id: null, ...thread}, options);
-  logger.info('[createAssistant] Updated thread:', updatedThread);
-
-  return thread;
+  throw new Error(`Assistant with ID "${assistant_id}" not found`);
 };
 
-const addMsgAssistant = async (code, inMessages, role = 'user', reply = false, skipSystemMessage = false) => {
-  try {
-    let thread = await Thread.findOne({ code: code });
-    logger.info(thread);
-    if (thread === null) return null;
-
-    const provider = createProvider({ variant: process.env.VARIANT || 'assistants' });
-    
-    await withThreadRecovery(
-      async (recoveredThread = thread) => {
-        thread = recoveredThread;
-        for (const message of inMessages) {
-          logger.info(message);
-          await provider.addMessage({
-            threadId: thread.getConversationId(),
-            role: role,
-            content: message
-          });
-          
-          // Save system message to database for frontend visibility
-          // Skip if message is already saved (e.g., from getConversationReplyController)
-          if (!skipSystemMessage) {
-            try {
-              const message_id = `system_${Date.now()}_${Math.random().toString(36).substring(7)}`;
-              await insertMessage({
-                nombre_whatsapp: 'System',
-                numero: code,
-                body: message,
-                timestamp: new Date(),
-                message_id: message_id,
-                is_group: false,
-                is_media: false,
-                from_me: true,
-                processed: true,
-                origin: 'system',
-                thread_id: thread.getConversationId(),
-                assistant_id: thread.getAssistantId(),
-                raw: { role: role }
-              });
-            } catch (err) {
-              // Don't throw - we don't want to break the flow if logging fails
-              logger.error('[addMsgAssistant] Error saving system message:', err);
-            }
-          }
-        }
-      },
-      thread,
-      process.env.VARIANT || 'assistants'
-    );
-
-    if (!reply) return null;
-
-    let output, completed;
-    let retries = 0;
-    const maxRetries = DEFAULT_MAX_RETRIES;
-    const assistant = getAssistantById(thread.getAssistantId(), thread);
-    do {
-      ({ output, completed } = await runAssistantAndWait({ thread, assistant }));
-      logger.info(`Attempt ${retries + 1}: completed=${completed}, output=${output || '(empty)'}`);
-      
-      if (completed && output) break;
-      if (retries < maxRetries) await new Promise(resolve => setTimeout(resolve, 2000));
-      retries++;
-    } while (retries <= maxRetries && (!completed || !output));
-
-    logger.info('THE ANS IS', output);
-    return output;
-  } catch (error) {
-    logger.info(error);
-    return null;
-  }
-};
-
-const addInstructionCore = async (code, instruction, role = 'user') => {
-  const thread = await withTracing(getThread, 'get_thread_operation', 
-    (threadCode) => ({ 
-      'thread.code': threadCode,
-      'operation.type': 'thread_retrieval' 
-    })
-  )(code);
-  if (thread === null) return null;
-
-  const assistant = getAssistantById(thread.getAssistantId(), thread);
-  const { output, completed, retries } = await withTracing(
-    runAssistantWithRetries,
-    'run_assistant_with_retries',
-    (thread, assistant, runConfig, patientReply) => ({
-      'assistant.id': thread.getAssistantId(),
-      'assistant.max_retries': DEFAULT_MAX_RETRIES,
-      'assistant.has_patient_reply': !!patientReply
-    })
-  )(
-    thread,
-    assistant,
-    {
-      additionalInstructions: instruction,
-      additionalMessages: [
-        { role: role, content: instruction }
-      ]
-    },
-    null // no patientReply for instructions
-  );
-
-  // Save instruction to database for frontend visibility
-  try {
-    const message_id = `instruction_${Date.now()}_${Math.random().toString(36).substring(7)}`;
-    await insertMessage({
-      nombre_whatsapp: 'Instruction',
-      numero: code,
-      body: instruction,
-      timestamp: new Date(),
-      message_id: message_id,
-      is_group: false,
-      is_media: false,
-      from_me: true,
-      processed: true,
-      origin: 'instruction',
-      thread_id: thread.getConversationId(),
-      assistant_id: thread.getAssistantId(),
-      raw: { role: role }
-    });
-  } catch (err) {
-    // Don't throw - we don't want to break the flow if logging fails
-    logger.error('[addInstructionCore] Error saving instruction message:', err);
-  }
+const createAssistant = withTracing(
+  (code, assistant_id) => createAssistantCore(code, assistant_id, getAssistantById),
+  'create_assistant',
+  (code, assistant_id) => ({
+    'assistant.thread_code': code,
+    'assistant.id': assistant_id,
+    'operation.type': 'create_assistant'
+  })
+);
 
-  logger.info('RUN RESPONSE', output);
-  return output;
-};
+const addMsgAssistant = withTracing(
+  addMsgAssistantCore,
+  'add_message_assistant',
+  (code, message, role) => ({
+    'message.thread_code': code,
+    'message.content_length': message?.length || 0,
+    'message.role': role,
+    'operation.type': 'add_message'
+  })
+);
 
 const addInsAssistant = withTracing(
   addInstructionCore,
@@ -329,226 +137,28 @@ const addInsAssistant = withTracing(
   })
 );
 
-const startTypingIndicator = async (provider, code) => {
-  try {
-    const lastMessage = await Message.findOne({ 
-      numero: code, 
-      from_me: false,
-      message_id: { $exists: true, $ne: null }
-    }).sort({ createdAt: -1 });
-    
-    if (!lastMessage?.message_id) return null;
-    
-    return setInterval(() => 
-      provider.sendTypingIndicator(lastMessage.message_id).catch(err => 
-        logger.debug('[startTypingIndicator] Interval failed', { error: err.message })
-      ), 25000
-    );
-  } catch (err) {
-    logger.debug('[startTypingIndicator] Failed to start', { error: err.message });
-    return null;
-  }
-};
-
-const replyAssistantCore = async (code, message_ = null, thread_ = null, runOptions = {}) => {
-  const timings = {};
-  const startTotal = Date.now();
-  
-  const provider = createProvider({ variant: process.env.VARIANT || 'assistants' });
-  const typingInterval = await startTypingIndicator(provider, code);
-  
-  try {
-    const { result: thread, duration: getThreadMs } = await withTracing(
-      getThread, 
-      'get_thread_operation', 
-      (threadCode) => ({ 
-        'thread.code': threadCode,
-        'operation.type': 'thread_retrieval',
-        'thread.provided': !!thread_
-      }),
-      { returnTiming: true }
-    )(code);
-    timings.get_thread_ms = getThreadMs;
-    
-    if (!thread_ && !thread) return null;
-    const finalThread = thread_ || thread;
-
-    const { result: patientReply, duration: getMessagesMs } = await withTracing(
-      getLastMessages,
-      'get_last_messages',
-      (code) => ({ 'thread.code': code }),
-      { returnTiming: true }
-    )(code);
-    timings.get_messages_ms = getMessagesMs;
-    
-    if (!patientReply) {
-      logger.info('[replyAssistantCore] No relevant data found for this assistant.');
-      return null;
-    }
-  
-    logger.info(`[replyAssistantCore] Processing ${patientReply.length} messages in parallel`);
-    const { result: processResult, duration: processMessagesMs } = await withTracing(
-      processThreadMessage,
-      'process_thread_messages',
-      (code, patientReply, provider) => ({
-        'messages.count': patientReply.length,
-        'thread.code': code
-      }),
-      { returnTiming: true }
-    )(code, patientReply, provider);
-    
-    const { results: processResults, timings: processTimings } = processResult;
-    timings.process_messages_ms = processMessagesMs;
-    
-    logger.debug('[replyAssistantCore] Process timings breakdown', { processTimings });
-  
-    if (processTimings) {
-      timings.process_messages_breakdown = {
-        download_ms: processTimings.download_ms || 0,
-        image_analysis_ms: processTimings.image_analysis_ms || 0,
-        audio_transcription_ms: processTimings.audio_transcription_ms || 0,
-        url_generation_ms: processTimings.url_generation_ms || 0,
-        total_media_ms: processTimings.total_media_ms || 0
-      };
-    }
-  
-    const patientMsg = processResults.some(r => r.isPatient);
-    const urls = processResults.filter(r => r.url).map(r => ({ url: r.url }));
-    const allMessagesToAdd = processResults.flatMap(r => r.messages || []);
-    const allTempFiles = processResults.flatMap(r => r.tempFiles || []);
-    
-    if (allMessagesToAdd.length > 0) {
-      logger.info(`[replyAssistantCore] Adding ${allMessagesToAdd.length} messages to thread in batch`);
-      await withThreadRecovery(
-        async (thread = finalThread) => {
-          const threadId = thread.getConversationId();
-          await provider.addMessage({ threadId, messages: allMessagesToAdd });
-        },
-        finalThread,
-        process.env.VARIANT || 'assistants'
-      );
-    }
-  
-    await Promise.all(processResults.map(r => updateMessageRecord(r.reply, finalThread)));
-    await cleanupFiles(allTempFiles);
-  
-    if (urls.length > 0) {
-      logger.info(`[replyAssistantCore] Processing ${urls.length} URLs for PDF combination`);
-      const { result: pdfResult, duration: pdfCombinationMs } = await withTracing(
-        combineImagesToPDF,
-        'combine_images_to_pdf',
-        ({ code }) => ({
-          'pdf.thread_code': code,
-          'pdf.url_count': urls.length
-        }),
-        { returnTiming: true }
-      )({ code });
-      timings.pdf_combination_ms = pdfCombinationMs;
-      const { pdfBuffer, processedFiles } = pdfResult;
-      logger.info(`[replyAssistantCore] PDF combination complete: ${processedFiles?.length || 0} files processed`);
-      
-      if (pdfBuffer) {
-        const key = `${code}-${Date.now()}-combined.pdf`;
-        const bucket = runtimeConfig.get('AWS_S3_BUCKET_NAME');
-        if (bucket) {
-          await AWS.uploadBufferToS3(pdfBuffer, bucket, key, 'application/pdf');
-        }
-      }
-      
-      if (processedFiles && processedFiles.length) {
-        cleanupFiles(processedFiles);
-      }
-    }
-
-    if (!patientMsg || finalThread.stopped) return null;
-
-    const assistant = getAssistantById(finalThread.getAssistantId(), finalThread);
-    const { result: runResult, duration: runAssistantMs } = await withTracing(
-      runAssistantWithRetries,
-      'run_assistant_with_retries',
-      (thread, assistant, runConfig, patientReply) => ({
-        'assistant.id': thread.getAssistantId(),
-        'assistant.max_retries': DEFAULT_MAX_RETRIES,
-        'assistant.has_patient_reply': !!patientReply
-      }),
-      { returnTiming: true }
-    )(finalThread, assistant, runOptions, patientReply);
-    timings.run_assistant_ms = runAssistantMs;
-    timings.total_ms = Date.now() - startTotal;
-    
-    const { run, output, completed, retries, predictionTimeMs, tools_executed } = runResult;
-
-    logger.info('[Assistant Reply Complete]', {
-      code: code ? `${code.substring(0, 3)}***${code.slice(-4)}` : 'unknown',
-      messageCount: patientReply.length,
-      hasMedia: urls.length > 0,
-      retries,
-      totalMs: timings.total_ms,
-      toolsExecuted: tools_executed?.length || 0
-    });
-
-    if (output && predictionTimeMs) {
-      logger.debug('[replyAssistantCore] Storing metrics with timing_breakdown', { 
-        timing_breakdown: timings,
-        has_breakdown: !!timings.process_messages_breakdown
-      });
-      
-      await PredictionMetrics.create({
-        message_id: `${code}-${Date.now()}`,
-        numero: code,
-        assistant_id: finalThread.getAssistantId(),
-        thread_id: finalThread.getConversationId(),
-        prediction_time_ms: predictionTimeMs,
-        retry_count: retries,
-        completed: completed,
-        timing_breakdown: timings
-      }).catch(err => logger.error('[replyAssistantCore] Failed to store metrics:', err));
-    }
-
-    return { output, tools_executed };
-  } finally {
-    if (typingInterval) {
-      clearInterval(typingInterval);
-    }
-  }
-};
-
 const replyAssistant = withTracing(
-  replyAssistantCore,
+  (code, message_, thread_, runOptions) => replyAssistantCore(code, message_, thread_, runOptions, getAssistantById),
   'assistant_reply',
   (code, message_, thread_, runOptions) => ({
     'assistant.thread_code': code,
     'assistant.has_message': !!message_,
     'assistant.has_custom_thread': !!thread_,
-    'assistant.run_options': JSON.stringify(runOptions)
+    'assistant.has_run_options': !!runOptions && Object.keys(runOptions).length > 0
   })
 );
 
-const switchAssistant = async (code, assistant_id) => {
-  try {
-    const thread = await Thread.findOne({ code: code });
-    logger.info('Inside thread', thread);
-    if (thread === null) return;
-
-    const variant = process.env.VARIANT || 'assistants';
-    const updateFields = { active: true, stopped: false };
-    
-    if (variant === 'responses') {
-      updateFields.prompt_id = assistant_id;
-    } else {
-      updateFields.assistant_id = assistant_id;
-    }
-
-    await Thread.updateOne({ code }, { $set: updateFields });  
-  } catch (error) {
-    logger.info(error);
-    return null;
-  }
-};
+const switchAssistant = withTracing(
+  switchAssistantCore,
+  'switch_assistant',
+  (code, assistant_id) => ({
+    'assistant.thread_code': code,
+    'assistant.new_id': assistant_id,
+    'operation.type': 'switch_assistant'
+  })
+);
 
 module.exports = {
-  getThread,
-  getThreadInfo,
   getAssistantById,
   createAssistant,
   replyAssistant,
@@ -557,6 +167,5 @@ module.exports = {
   switchAssistant,
   configureAssistants,
   registerAssistant,
-  overrideGetAssistantById,
-  runAssistantAndWait
+  overrideGetAssistantById
 };

package/lib/services/assistantServiceCore.js

@@ -0,0 +1,215 @@
+const AWS = require('../config/awsConfig.js');
+const runtimeConfig = require('../config/runtimeConfig');
+const { createProvider } = require('../providers/createProvider');
+
+const { Thread } = require('../models/threadModel.js');
+const { PredictionMetrics } = require('../models/predictionMetricsModel');
+const { insertMessage } = require('../models/messageModel');
+
+const { getCurRow, runAssistantWithRetries } = require('../helpers/assistantHelper.js');
+const { getThread } = require('../helpers/threadHelper.js');
+const { processThreadMessage } = require('../helpers/processHelper.js');
+const { getLastMessages, updateMessageRecord } = require('../helpers/messageHelper.js');
+const { withThreadRecovery } = require('../helpers/threadRecoveryHelper.js');
+const { combineImagesToPDF, cleanupFiles } = require('../helpers/filesHelper.js');
+const { logger } = require('../utils/logger');
+
+const createAssistantCore = async (code, assistant_id, getAssistantById) => {
+  const thread = await getThread(code);
+  if (!thread) return null;
+
+  const assistant = getAssistantById(assistant_id, thread);
+  const curRow = await getCurRow(code);
+  const context = { curRow };
+
+  try {
+    await assistant.create(code, context);
+    return { success: true, assistant_id };
+  } catch (error) {
+    logger.error('[createAssistantCore] Error:', error);
+    return { success: false, error: error.message };
+  }
+};
+
+const addMsgAssistantCore = async (code, message, role = 'user') => {
+  const thread = await getThread(code);
+  if (!thread) return null;
+
+  const provider = createProvider({ variant: process.env.VARIANT || 'assistants' });
+  const threadId = thread.getConversationId();
+
+  try {
+    await provider.addMessage({ threadId, messages: [{ role, content: message }] });
+    await insertMessage({ code, message, role });
+    return { success: true };
+  } catch (error) {
+    logger.error('[addMsgAssistantCore] Error:', error);
+    return { success: false, error: error.message };
+  }
+};
+
+const addInstructionCore = async (code, instruction, role = 'user') => {
+  const thread = await getThread(code);
+  if (!thread) return null;
+
+  const provider = createProvider({ variant: process.env.VARIANT || 'assistants' });
+  const threadId = thread.getConversationId();
+
+  try {
+    await provider.addMessage({ threadId, messages: [{ role, content: instruction }] });
+    return { success: true };
+  } catch (error) {
+    logger.error('[addInstructionCore] Error:', error);
+    return { success: false, error: error.message };
+  }
+};
+
+const replyAssistantCore = async (code, message_ = null, thread_ = null, runOptions = {}, getAssistantById) => {
+  const timings = {};
+  const startTotal = Date.now();
+  
+  try {
+    const thread = thread_ || await getThread(code);
+    timings.get_thread_ms = 0;
+    
+    if (!thread) return null;
+    const finalThread = thread;
+
+    const patientReply = await getLastMessages(code);
+    timings.get_messages_ms = 0;
+    
+    if (!patientReply) {
+      logger.info('[replyAssistantCore] No relevant data found for this assistant.');
+      return null;
+    }
+
+    const provider = createProvider({ variant: process.env.VARIANT || 'assistants' });
+    logger.info(`[replyAssistantCore] Processing ${patientReply.length} messages in parallel`);
+    const processResult = await processThreadMessage(code, patientReply, provider);
+    
+    const { results: processResults, timings: processTimings } = processResult;
+    timings.process_messages_ms = 0;
+    
+    logger.debug('[replyAssistantCore] Process timings breakdown', { processTimings });
+
+    if (processTimings) {
+      timings.process_messages_breakdown = {
+        download_ms: processTimings.download_ms || 0,
+        image_analysis_ms: processTimings.image_analysis_ms || 0,
+        audio_transcription_ms: processTimings.audio_transcription_ms || 0,
+        url_generation_ms: processTimings.url_generation_ms || 0,
+        total_media_ms: processTimings.total_media_ms || 0
+      };
+    }
+
+    const patientMsg = processResults.some(r => r.isPatient);
+    const urls = processResults.filter(r => r.url).map(r => ({ url: r.url }));
+    const allMessagesToAdd = processResults.flatMap(r => r.messages || []);
+    const allTempFiles = processResults.flatMap(r => r.tempFiles || []);
+    
+    if (allMessagesToAdd.length > 0) {
+      logger.info(`[replyAssistantCore] Adding ${allMessagesToAdd.length} messages to thread in batch`);
+      await withThreadRecovery(
+        async (thread = finalThread) => {
+          const threadId = thread.getConversationId();
+          await provider.addMessage({ threadId, messages: allMessagesToAdd });
+        },
+        finalThread,
+        process.env.VARIANT || 'assistants'
+      );
+    }
+
+    await Promise.all(processResults.map(r => updateMessageRecord(r.reply, finalThread)));
+    await cleanupFiles(allTempFiles);
+
+    if (urls.length > 0) {
+      logger.info(`[replyAssistantCore] Processing ${urls.length} URLs for PDF combination`);
+      const pdfResult = await combineImagesToPDF({ code });
+      timings.pdf_combination_ms = 0;
+      const { pdfBuffer, processedFiles } = pdfResult;
+      logger.info(`[replyAssistantCore] PDF combination complete: ${processedFiles?.length || 0} files processed`);
+      
+      if (pdfBuffer) {
+        const key = `${code}-${Date.now()}-combined.pdf`;
+        const bucket = runtimeConfig.get('AWS_S3_BUCKET_NAME');
+        if (bucket) {
+          await AWS.uploadBufferToS3(pdfBuffer, bucket, key, 'application/pdf');
+        }
+      }
+      
+      if (processedFiles && processedFiles.length) {
+        cleanupFiles(processedFiles);
+      }
+    }
+
+    if (!patientMsg || finalThread.stopped) return null;
+
+    const assistant = getAssistantById(finalThread.getAssistantId(), finalThread);
+    const runResult = await runAssistantWithRetries(finalThread, assistant, runOptions, patientReply);
+    timings.run_assistant_ms = 0;
+    timings.total_ms = Date.now() - startTotal;
+    
+    const { run, output, completed, retries, predictionTimeMs, tools_executed } = runResult;
+
+    logger.info('[Assistant Reply Complete]', {
+      code: code ? `${code.substring(0, 3)}***${code.slice(-4)}` : 'unknown',
+      messageCount: patientReply.length,
+      hasMedia: urls.length > 0,
+      retries,
+      totalMs: timings.total_ms,
+      toolsExecuted: tools_executed?.length || 0
+    });
+
+    if (output && predictionTimeMs) {
+      logger.debug('[replyAssistantCore] Storing metrics with timing_breakdown', { 
+        timing_breakdown: timings,
+        has_breakdown: !!timings.process_messages_breakdown
+      });
+      
+      await PredictionMetrics.create({
+        message_id: `${code}-${Date.now()}`,
+        numero: code,
+        assistant_id: finalThread.getAssistantId(),
+        thread_id: finalThread.getConversationId(),
+        prediction_time_ms: predictionTimeMs,
+        retry_count: retries,
+        completed: completed,
+        timing_breakdown: timings
+      }).catch(err => logger.error('[replyAssistantCore] Failed to store metrics:', err));
+    }
+
+    return { output, tools_executed };
+  } catch (error) {
+    logger.error('[replyAssistantCore] Error:', { error: error.message });
+    throw error;
+  }
+};
+
+const switchAssistantCore = async (code, assistant_id) => {
+  try {
+    const thread = await Thread.findOne({ code });
+    if (!thread) {
+      return null;
+    }
+
+    const updateFields = {
+      assistant_id,
+      stopped: false,
+      updatedAt: new Date()
+    };
+
+    await Thread.updateOne({ code }, { $set: updateFields });
+    return { success: true, assistant_id };
+  } catch (error) {
+    logger.info(error);
+    return null;
+  }
+};
+
+module.exports = {
+  createAssistantCore,
+  addMsgAssistantCore,
+  addInstructionCore,
+  replyAssistantCore,
+  switchAssistantCore
+};

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@peopl-health/nexus",
-  "version": "2.5.1",
+  "version": "2.5.2",
   "description": "Core messaging and assistant library for WhatsApp communication platforms",
   "keywords": [
     "whatsapp",