feat: Implement structured message parsing and display in MessageViewer component

- Added MessageViewer component to display parsed messages in a chat-like format.
- Introduced new Message table in the database to store individual messages with timestamps, roles, and content.
- Updated Session model to include a relation to parsed messages.
- Created transcript parsing logic to convert raw transcripts into structured messages.
- Enhanced processing scheduler to handle sessions with parsed messages.
- Updated API endpoints to return parsed messages alongside session details.
- Added manual trigger commands for session refresh, transcript parsing, and processing.
- Improved user experience with color-coded message roles and timestamps in the UI.
- Documented the new scheduler workflow and transcript parsing implementation.

lib/processingScheduler.js

@@ -181,46 +181,59 @@ function validateOpenAIResponse(data) {
 /**
  * Process unprocessed sessions
  */
-async function processUnprocessedSessions() {
+export async function processUnprocessedSessions() {
   process.stdout.write("[ProcessingScheduler] Starting to process unprocessed sessions...\n");
 
-  // Find sessions that have transcript content but haven't been processed
+  // Find sessions that have messages but haven't been processed
   const sessionsToProcess = await prisma.session.findMany({
     where: {
       AND: [
-        { transcriptContent: { not: null } },
-        { transcriptContent: { not: "" } },
         { processed: { not: true } }, // Either false or null
       ],
     },
-    select: {
-      id: true,
-      transcriptContent: true,
+    include: {
+      messages: {
+        orderBy: { order: 'asc' }
+      }
     },
     take: 10, // Process in batches to avoid overloading the system
   });
 
-  if (sessionsToProcess.length === 0) {
+  // Filter to only sessions that have messages
+  const sessionsWithMessages = sessionsToProcess.filter(session => session.messages.length > 0);
+
+  if (sessionsWithMessages.length === 0) {
     process.stdout.write("[ProcessingScheduler] No sessions found requiring processing.\n");
     return;
   }
 
-  process.stdout.write(`[ProcessingScheduler] Found ${sessionsToProcess.length} sessions to process.\n`);
+  process.stdout.write(`[ProcessingScheduler] Found ${sessionsWithMessages.length} sessions to process.\n`);
   let successCount = 0;
   let errorCount = 0;
 
-  for (const session of sessionsToProcess) {
-    if (!session.transcriptContent) {
-      // Should not happen due to query, but good for type safety
-      process.stderr.write(`[ProcessingScheduler] Session ${session.id} has no transcript content, skipping.\n`);
+  for (const session of sessionsWithMessages) {
+    if (session.messages.length === 0) {
+      process.stderr.write(`[ProcessingScheduler] Session ${session.id} has no messages, skipping.\n`);
       continue;
     }
 
-    process.stdout.write(`[ProcessingScheduler] Processing transcript for session ${session.id}...\n`);
+    process.stdout.write(`[ProcessingScheduler] Processing messages for session ${session.id}...\n`);
     try {
+      // Convert messages back to transcript format for OpenAI processing
+      const transcript = session.messages.map(msg =>
+        `[${new Date(msg.timestamp).toLocaleString('en-GB', {
+          day: '2-digit',
+          month: '2-digit',
+          year: 'numeric',
+          hour: '2-digit',
+          minute: '2-digit',
+          second: '2-digit'
+        }).replace(',', '')}] ${msg.role}: ${msg.content}`
+      ).join('\n');
+
       const processedData = await processTranscriptWithOpenAI(
         session.id,
-        session.transcriptContent
+        transcript
       );
 
       // Map sentiment string to float value for compatibility with existing data