gradio logging

gradio/gradio/chatbot/config.py

@@ -20,8 +20,27 @@ def get_config(key, default=None):
     return value
 
 def get_logger(name):
-    log = logging.getLogger(name)
-    return log
+    class Logger:
+        def __init__(self, name):
+            self.name = name
+
+        def info(self, message):
+            print(f"[INFO] {self.name}: {message}")
+            sys.stdout.flush()
+
+        def error(self, message):
+            print(f"[ERROR] {self.name}: {message}", file=sys.stderr)
+            sys.stderr.flush()
+
+        def warning(self, message):
+            print(f"[WARNING] {self.name}: {message}")
+            sys.stdout.flush()
+
+        def debug(self, message):
+            print(f"[DEBUG] {self.name}: {message}")
+            sys.stdout.flush()
+
+    return Logger(name)
 
 def invalid_config(variable, error_message):
     "Log error for invalid configuration and quit"

gradio/gradio/chatbot/main.py

@@ -8,6 +8,9 @@
 
 # Function to stream responses from the LLM, maintaining context
 def query_llm(input_text, history):
+    # Log the user input
+    log.debug(f"User Input: {input_text}")
+
     # Include the entire conversation history in the payload
     messages = [{"role": "user", "content": user_message} if i % 2 == 0 else {"role": "assistant", "content": assistant_message}
                 for i, (user_message, assistant_message) in enumerate(history)]
@@ -27,35 +30,40 @@ def query_llm(input_text, history):
 
     # Check if the response is valid
     if response.status_code == 200:
-        partial_response = ""
-        
+        full_response = ""
+
         # Stream the response in chunks
         for line in response.iter_lines():
             if line:
                 # Decode and load the JSON from each 'data:' chunk
                 line_str = line.decode('utf-8')
-                
+
                 # Skip the DONE message
                 if line_str.strip() == "data: [DONE]":
                     break
-                
+
                 if line_str.startswith("data:"):
                     data_json = line_str[6:]  # Remove 'data: ' prefix
                     data = json.loads(data_json)
-                    
+
                     # Extract and append content from the "delta"
                     delta_content = data["choices"][0]["delta"].get("content", "")
-                    partial_response += delta_content
-                    
+                    full_response += delta_content
+
                     # Yield the partial response so far
-                    yield partial_response
+                    yield full_response
+
+        # Log the complete response after all chunks are received
+        log.debug(f"Assistant Full Response: {full_response}")
     else:
-        yield f"Error: {response.status_code}"
+        error_message = f"Error: {response.status_code}"
+        log.error(error_message)
+        yield error_message
 
 # Using gr.ChatInterface to handle streaming
 chatbot = gr.ChatInterface(fn=query_llm, title="LM Studio Chat")
 
-print("Launching gradio interface ...")
+print("Launching Gradio interface ...")
 sys.stdout.flush()
-    
+
 chatbot.launch(server_name="0.0.0.0", server_port=7860)