ts: file upload max context

app/nodes.py

@@ -198,18 +198,29 @@ async def generate_node_streaming(state: GraphState) -> Generator[GraphState, No
     try:
         # Get MAX_CONTEXT_CHARS at the beginning so it's available throughout the function
         MAX_CONTEXT_CHARS = int(config.get("general", "MAX_CONTEXT_CHARS"))
+        logger.info(f"Using MAX_CONTEXT_CHARS: {MAX_CONTEXT_CHARS}")
         
         # Combine retriever context with ingestor context
         retrieved_context = state.get("context", "")
         ingestor_context = state.get("ingestor_context", "")
         
+        logger.info(f"Original context lengths - Ingestor: {len(ingestor_context)}, Retrieved: {len(retrieved_context)}")
+        
         # Convert contexts to list format expected by generator
         context_list = []
+        total_context_chars = 0
         
         if ingestor_context:
+            # Truncate ingestor context if it's too long
+            if len(ingestor_context) > MAX_CONTEXT_CHARS:
+                logger.warning(f"Truncating ingestor context from {len(ingestor_context)} to {MAX_CONTEXT_CHARS} characters")
+                truncated_ingestor = ingestor_context[:MAX_CONTEXT_CHARS] + "...\n[Content truncated due to length]"
+            else:
+                truncated_ingestor = ingestor_context
+            
             # Add ingestor context
             context_list.append({
-                "answer": ingestor_context,
+                "answer": truncated_ingestor,
                 "answer_metadata": {
                     "filename": state.get("filename", "Uploaded Document"),
                     "page": "Unknown",
@@ -217,11 +228,29 @@ async def generate_node_streaming(state: GraphState) -> Generator[GraphState, No
                     "source": "Ingestor"
                 }
             })
+            total_context_chars += len(truncated_ingestor)
         
-        if retrieved_context:
+        if retrieved_context and total_context_chars < MAX_CONTEXT_CHARS:
             # Convert retrieved context to list and add
             retrieved_list = convert_context_to_list(retrieved_context)
-            context_list.extend(retrieved_list)
+            
+            # Add retrieved context items until we hit the limit
+            remaining_chars = MAX_CONTEXT_CHARS - total_context_chars
+            for item in retrieved_list:
+                item_text = item.get("answer", "")
+                if len(item_text) <= remaining_chars:
+                    context_list.append(item)
+                    remaining_chars -= len(item_text)
+                else:
+                    # Truncate this item and stop
+                    if remaining_chars > 100:  # Only add if we have meaningful space left
+                        item["answer"] = item_text[:remaining_chars-50] + "...\n[Content truncated]"
+                        context_list.append(item)
+                    break
+        
+        # Calculate final context size
+        final_context_size = sum(len(item.get("answer", "")) for item in context_list)
+        logger.info(f"Final context size: {final_context_size} characters (limit: {MAX_CONTEXT_CHARS})")
         
         # Prepare the request payload
         payload = {
@@ -285,7 +314,8 @@ async def generate_node_streaming(state: GraphState) -> Generator[GraphState, No
                                             "result_length": len(current_text),
                                             "generation_success": True,
                                             "streaming": True,
-                                            "generator_type": "fastapi"
+                                            "generator_type": "fastapi",
+                                            "context_chars_used": final_context_size
                                         })
                                         
                                         yield {
@@ -302,7 +332,8 @@ async def generate_node_streaming(state: GraphState) -> Generator[GraphState, No
                                         "result_length": len(current_text),
                                         "generation_success": True,
                                         "streaming": True,
-                                        "generator_type": "fastapi"
+                                        "generator_type": "fastapi",
+                                        "context_chars_used": final_context_size
                                     })
                                     
                                     yield {