Production

Channel-Backend/src/services/chat.service.ts

@@ -92,12 +92,17 @@ export class ChatService {
     }
 
     // 2. Fetch Embeddings & Data Sources
-    let embeddingsCount = await prisma.assetEmbedding.count({
-      where: { assetId: { in: accessibleAssetIds } }
+    // Ensure ALL accessible assets have embeddings generated
+    const embeddedAssets = await prisma.assetEmbedding.findMany({
+      where: { assetId: { in: accessibleAssetIds } },
+      select: { assetId: true },
+      distinct: ['assetId']
     });
+    const embeddedAssetIds = embeddedAssets.map(e => e.assetId);
+    const missingAssetIds = accessibleAssetIds.filter(id => !embeddedAssetIds.includes(id));
 
-    if (embeddingsCount === 0 && accessibleAssetIds.length > 0) {
-      await this.autoIndexCatalog(accessibleAssetIds);
+    if (missingAssetIds.length > 0) {
+      await this.autoIndexCatalog(missingAssetIds);
     }
 
     const embeddings = await prisma.assetEmbedding.findMany({

Channel-Backend/src/services/extraction.service.ts

@@ -266,23 +266,26 @@ export class ExtractionService {
           }
         });
       }
-      // E. Plain Text / Markdown
-      else {
+      // E. Plain Text / Markdown / Code (Excluding binaries)
+      else if (!ext.match(/\.(png|jpe?g|gif|webp|svg|mp4|webm|avi|mp3|wav)$/i)) {
         const textContent = buffer.toString('utf-8');
-        const subChunks = this.splitText(textContent, 600);
-        subChunks.forEach((text, i) => {
-          chunks.push({
-            chunkIndex: chunkIndex++,
-            chunkType: 'TEXT',
-            content: text,
-            sourceMetadata: {
-              assetId: asset.id,
-              assetTitle: asset.title,
-              assetType: asset.type,
-              location: `Document Text: Segment ${i + 1}`,
-            },
+        // Only proceed if it looks like actual text (not arbitrary binary data)
+        if (!textContent.includes('\u0000\u0000')) {
+          const subChunks = this.splitText(textContent, 600);
+          subChunks.forEach((text, i) => {
+            chunks.push({
+              chunkIndex: chunkIndex++,
+              chunkType: 'TEXT',
+              content: text,
+              sourceMetadata: {
+                assetId: asset.id,
+                assetTitle: asset.title,
+                assetType: asset.type,
+                location: `Document Text: Segment ${i + 1}`,
+              },
+            });
           });
-        });
+        }
       }
     }