Update app.py

app.py

@@ -113,27 +113,7 @@ def smart_truncate(text, max_length=3000):
     if len(sentences) > 1:
         return ' '.join(sentences[:-1]) + "... [Response truncated - ask for continuation]"
     # Otherwise, split by word
-    else:
-        words = text[:max_length].split()
-        return ' '.join(words[:-1]) + "... [Response truncated - ask for continuation]"
-
-def detect_tool_request(text):
-    """Simple heuristic to detect when a graph might be helpful."""
-    graph_keywords = [
-        "chart", "graph", "plot", "visualize", "show data", "bar chart", 
-        "line graph", "pie chart", "diagram", "compare", "data visualization"
-    ]
-    text_lower = text.lower()
-    return any(keyword in text_lower for keyword in graph_keywords)
-
-def call_hf_api(messages, max_retries=3):
-    """Call Hugging Face API with retry logic."""
-    payload = {
-        "inputs": format_messages_for_hf(messages),
-        "parameters": {
-            "max_new_tokens": 1024,
-            "temperature": 0.7,
-            "top_p": 0.9,
++            "top_p": 0.9,
             "return_full_text": False
         }
     }
@@ -237,69 +217,59 @@ def process_response_for_tools(response_text, original_query):
     return response_text
 
 def respond_with_enhanced_streaming(message, history):
-    """Generate response using HF API with tool support."""
+    """Streams the bot's response, with support for graph generation."""
     timing_context = metrics_tracker.start_timing()
     error_occurred = False
-    error_message = None
-    full_response = ""
-    
+    response = ""
+    mode = ""
+
     try:
-        # Prepare messages for API
-        api_messages = [{"role": "system", "content": SYSTEM_MESSAGE}]
-        
-        if history:
-            # Handle message history
-            for exchange in history[-5:]:  # Use last 5 exchanges
-                if isinstance(exchange, dict):
-                    api_messages.append(exchange)
-                else:
-                    # Fallback for other formats
-                    api_messages.append({"role": "user", "content": str(exchange[0])})
-                    api_messages.append({"role": "assistant", "content": str(exchange[1])})
-        
-        api_messages.append({"role": "user", "content": message})
-        
-        metrics_tracker.mark_provider_start(timing_context)
-        
-        # Get response from HF API
-        response_text = call_hf_api(api_messages)
-        
-        # Process response for potential tool usage
-        response_text = process_response_for_tools(response_text, message)
-        
-        metrics_tracker.mark_provider_end(timing_context)
-        
-        # Simulate streaming by yielding chunks
-        words = response_text.split()
-        current_response = ""
-        
-        for i, word in enumerate(words):
-            current_response += word + " "
-            if i % 3 == 0:  # Yield every 3 words to simulate streaming
-                yield current_response.strip()
-                time.sleep(0.01)  # Small delay to simulate streaming
-        
-        # Final yield with complete response
-        full_response = current_response.strip()
-        yield full_response
-        
-        logger.info(f"Response completed. Length: {len(full_response)} characters")
-        
+        # ---- Call HuggingFace Inference ----
+        payload = {
+            "inputs": message,
+            "parameters": {"return_full_text": False},
+        }
+        headers = {"Authorization": f"Bearer {HF_API_TOKEN}"}
+        api_response = requests.post(HF_API_URL, headers=headers, json=payload)
+        api_response.raise_for_status()
+        outputs = api_response.json()
+
+        # HuggingFace might return [{"generated_text": "..."}]
+        model_output = outputs[0].get("generated_text", "").strip()
+
+        # ---- Check if it's a tool call ----
+        try:
+            parsed = json.loads(model_output)
+            if isinstance(parsed, dict) and parsed.get("name") == "create_graph":
+                args = parsed.get("arguments", {})
+                graph_html = generate_plot(
+                    args.get("data_json", "{}"),
+                    args.get("labels_json", "[]"),
+                    args.get("plot_type", "bar"),
+                    args.get("title", "Untitled"),
+                    args.get("x_label", ""),
+                    args.get("y_label", "")
+                )
+                response = graph_html
+                mode = "graph"
+            else:
+                response = model_output
+                mode = "text"
+        except json.JSONDecodeError:
+            # Not JSON → just plain text
+            response = model_output
+            mode = "text"
+
     except Exception as e:
         error_occurred = True
-        error_message = str(e)
-        logger.exception("Error in response generation")
-        full_response = "Sorry, an error occurred while generating the response."
-        yield full_response
+        response = f"⚠️ Error: {str(e)}"
+        mode = "error"
 
     finally:
-        metrics_tracker.log_interaction(
-            query=message,
-            response=full_response,
-            timing_context=timing_context,
-            error_occurred=error_occurred,
-            error_message=error_message,
-        )
+        metrics_tracker.stop_timing(timing_context, error_occurred)
+
+    return response, mode
+
 
 # ===============================================================================
 # UI CONFIGURATION SECTION - ALL UI RELATED CODE CENTRALIZED HERE