Update Gradio_UI.py

Gradio_UI.py

@@ -18,110 +18,82 @@ import os
 import re
 import shutil
 from typing import Optional
+import tempfile # Added for PIL image saving
+from PIL import Image as PILImage # Added for PIL image handling
 
 from smolagents.agent_types import AgentAudio, AgentImage, AgentText, handle_agent_output_types
-from smolagents.agents import ActionStep, MultiStepAgent # Ensure MultiStepAgent is correctly referenced
+from smolagents.agents import ActionStep, MultiStepAgent 
 from smolagents.memory import MemoryStep
 from smolagents.utils import _is_package_available
+import gradio as gr # Ensure gradio is imported at the top level
 
 
-def pull_messages_from_step(
-    step_log: MemoryStep,
-):
-    """Extract ChatMessage objects from agent steps with proper nesting"""
-    import gradio as gr
-
+def pull_messages_from_step_dict(step_log: MemoryStep):
+    """Extract messages as dicts for Gradio type='messages' Chatbot"""
     if isinstance(step_log, ActionStep):
-        # Output the step number
-        step_number = f"Step {step_log.step_number}" if step_log.step_number is not None else ""
-        yield gr.ChatMessage(role="assistant", content=f"**{step_number}**")
+        step_number_str = f"Step {step_log.step_number}" if step_log.step_number is not None else "Processing"
+        yield {"role": "assistant", "content": f"**{step_number_str}**"}
 
-        # First yield the thought/reasoning from the LLM
         if hasattr(step_log, "model_output") and step_log.model_output is not None:
-            # Clean up the LLM output
             model_output = step_log.model_output.strip()
-            # Remove any trailing <end_code> and extra backticks, handling multiple possible formats
-            model_output = re.sub(r"```\s*<end_code>", "```", model_output)  # handles ```<end_code>
-            model_output = re.sub(r"<end_code>\s*```", "```", model_output)  # handles <end_code>```
-            model_output = re.sub(r"```\s*\n\s*<end_code>", "```", model_output)  # handles ```\n<end_code>
+            # More robust cleaning for <end_code> potentially wrapped in backticks or with newlines
+            model_output = re.sub(r"```\s*<end_code>[\s\S]*|[\s\S]*<end_code>\s*```", "```", model_output, flags=re.DOTALL)
+            model_output = re.sub(r"<end_code>", "", model_output) # Remove standalone tag
             model_output = model_output.strip()
-            yield gr.ChatMessage(role="assistant", content=model_output)
-
-        # For tool calls, create a parent message
-        if hasattr(step_log, "tool_calls") and step_log.tool_calls is not None:
-            first_tool_call = step_log.tool_calls[0]
-            used_code = first_tool_call.name == "python_interpreter"
-            parent_id = f"call_{len(step_log.tool_calls)}_{step_log.step_number or 'x'}" # Make parent_id more unique
+            yield {"role": "assistant", "content": model_output}
 
-            # Tool call becomes the parent message with timing info
-            args = first_tool_call.arguments
+        if hasattr(step_log, "tool_calls") and step_log.tool_calls:
+            tc = step_log.tool_calls[0] # Process first tool call for simplicity in this format
+            tool_info_md = f"🛠️ **Tool Used: {tc.name}**\n"
+            
+            args = tc.arguments
             if isinstance(args, dict):
-                content = str(args.get("answer", str(args)))
+                args_str = str(args.get("answer", str(args)))
             else:
-                content = str(args).strip()
-
-            if used_code:
-                content = re.sub(r"```.*?\n", "", content)  # Remove existing code blocks
-                content = re.sub(r"\s*<end_code>\s*", "", content)  # Remove end_code tags
-                content = content.strip()
-                if not content.startswith("```python"): # Ensure it's a python block
-                    content = f"```python\n{content}\n```"
-                else: # If it is, ensure newlines are correct
-                    content = content.replace("```python", "```python\n").replace("\n```", "\n```")
-
-
-            parent_message_tool = gr.ChatMessage(
-                role="assistant",
-                content=content,
-                metadata={
-                    "title": f"🛠️ Used tool {first_tool_call.name}",
-                    "id": parent_id,
-                    "status": "pending",
-                },
-            )
-            yield parent_message_tool
-
-            if hasattr(step_log, "observations") and (
-                step_log.observations is not None and step_log.observations.strip()
-            ):
-                log_content = step_log.observations.strip()
-                if log_content: # Only yield if there's actual content
-                    log_content = re.sub(r"^Execution logs:\s*", "", log_content)
-                    yield gr.ChatMessage(
-                        role="assistant",
-                        content=f"{log_content}",
-                        metadata={"title": "📝 Execution Logs", "parent_id": parent_id, "status": "done"},
-                    )
+                args_str = str(args).strip()
+            
+            if tc.name == "python_interpreter":
+                code_content = args_str
+                # Clean up common wrapping issues
+                code_content = re.sub(r"^```python\s*\n?", "", code_content)
+                code_content = re.sub(r"\n?```\s*$", "", code_content)
+                code_content = re.sub(r"^\s*<end_code>\s*", "", code_content)
+                code_content = re.sub(r"\s*<end_code>\s*$", "", code_content)
+                code_content = code_content.strip()
+                tool_info_md += f"Executing Code:\n```python\n{code_content}\n```\n"
+            else:
+                tool_info_md += f"Arguments: `{args_str}`\n"
+
+            if hasattr(step_log, "observations") and step_log.observations and step_log.observations.strip():
+                obs_content = step_log.observations.strip()
+                # Remove "Execution logs:" prefix if present for cleaner display
+                obs_content = re.sub(r"^Execution logs:\s*", "", obs_content).strip()
+                if obs_content: # Only show if there's something after stripping
+                    tool_info_md += f"📝 **Tool Output/Logs:**\n```\n{obs_content}\n```\n"
+            
+            if hasattr(step_log, "error") and step_log.error:
+                tool_info_md += f"💥 **Error:** {str(step_log.error)}\n"
+            
+            yield {"role": "assistant", "content": tool_info_md.strip()}
 
-            if hasattr(step_log, "error") and step_log.error is not None:
-                yield gr.ChatMessage(
-                    role="assistant",
-                    content=str(step_log.error),
-                    metadata={"title": "💥 Error", "parent_id": parent_id, "status": "done"},
-                )
-            # This direct update might not work as expected as yield creates new objects.
-            # Status update is visual; actual logic might be more complex.
-            parent_message_tool.metadata["status"] = "done" 
-
-        elif hasattr(step_log, "error") and step_log.error is not None: # Standalone errors
-            yield gr.ChatMessage(role="assistant", content=str(step_log.error), metadata={"title": "💥 Error"})
-
-        step_footnote_parts = [step_number]
-        if hasattr(step_log, "input_token_count") and step_log.input_token_count is not None and \
-           hasattr(step_log, "output_token_count") and step_log.output_token_count is not None:
-            token_str = (
-                f" | Input-tokens:{step_log.input_token_count:,} | Output-tokens:{step_log.output_token_count:,}"
-            )
-            step_footnote_parts.append(token_str)
+        elif hasattr(step_log, "error") and step_log.error: # Standalone error not from a tool call
+            yield {"role": "assistant", "content": f"💥 **Error:** {str(step_log.error)}"}
+        
+        # --- Minimal footnote for type="messages" ---
+        footnote_parts = []
+        if step_log.step_number is not None:
+            footnote_parts.append(f"Step {step_log.step_number}")
         if hasattr(step_log, "duration") and step_log.duration is not None:
-            step_duration = f" | Duration: {round(float(step_log.duration), 2)}s"
-            step_footnote_parts.append(step_duration)
+            footnote_parts.append(f"Duration: {round(float(step_log.duration), 2)}s")
+        if hasattr(step_log, "input_token_count") and step_log.input_token_count is not None: # Check for None
+             footnote_parts.append(f"InTokens: {step_log.input_token_count:,}")
+        if hasattr(step_log, "output_token_count") and step_log.output_token_count is not None: # Check for None
+             footnote_parts.append(f"OutTokens: {step_log.output_token_count:,}")
         
-        step_footnote_text = "".join(filter(None, step_footnote_parts))
-        if step_footnote_text:
-             step_footnote = f"""<span style="color: #bbbbc2; font-size: 12px;">{step_footnote_text}</span> """
-             yield gr.ChatMessage(role="assistant", content=f"{step_footnote}")
-        yield gr.ChatMessage(role="assistant", content="-----")
+        if footnote_parts:
+            footnote_text = " | ".join(footnote_parts)
+            yield {"role": "assistant", "content": f"""<p style="color: #999; font-size: 0.8em; margin-top:0; margin-bottom:0;">{footnote_text}</p>"""}
+        yield {"role": "assistant", "content": "---"} # Separator
 
 
 def stream_to_gradio(
@@ -130,40 +102,65 @@ def stream_to_gradio(
     reset_agent_memory: bool = False,
     additional_args: Optional[dict] = None,
 ):
-    """Runs an agent with the given task and streams the messages from the agent as gradio ChatMessages."""
+    """Runs an agent, yields message dicts for Gradio type='messages' Chatbot."""
     if not _is_package_available("gradio"):
-        raise ModuleNotFoundError(
-            "Please install 'gradio' extra to use the GradioUI: `pip install 'smolagents[gradio]'`"
-        )
-    import gradio as gr
+        raise ModuleNotFoundError("Install 'gradio': `pip install 'smolagents[gradio]'`")
 
-    # Reset interaction logs for the new run if the agent has this attribute
-    if hasattr(agent, 'interaction_logs'):
+    if hasattr(agent, 'interaction_logs'): # Clear logs for this new agent run
         agent.interaction_logs.clear()
-        print("DEBUG: Cleared agent interaction_logs for new run.")
-
+        print("DEBUG Gradio: Cleared agent interaction_logs for new run.")
 
     for step_log in agent.run(task, stream=True, reset=reset_agent_memory, additional_args=additional_args):
-        if hasattr(agent.model, "last_input_token_count") and agent.model.last_input_token_count is not None: # Check for None
-            if isinstance(step_log, ActionStep): # Only add token counts to ActionSteps
+        if hasattr(agent.model, "last_input_token_count") and agent.model.last_input_token_count is not None:
+            if isinstance(step_log, ActionStep):
                 step_log.input_token_count = agent.model.last_input_token_count
                 step_log.output_token_count = agent.model.last_output_token_count
+        
+        for msg_dict in pull_messages_from_step_dict(step_log): # Use new dict-yielding function
+            yield msg_dict
 
-        for message in pull_messages_from_step(step_log):
-            yield message
+    final_answer_content = step_log # Last step_log is the final output/state
+
+    # --- Handle final answer for type="messages" ---
+    if isinstance(final_answer_content, PILImage.Image):
+        print("DEBUG Gradio (stream_to_gradio): Final answer is raw PIL Image.")
+        try:
+            with tempfile.NamedTemporaryFile(delete=False, suffix=".png") as tmp_file:
+                final_answer_content.save(tmp_file, format="PNG")
+                image_path_for_gradio = tmp_file.name
+            print(f"DEBUG Gradio: Saved PIL image to temp path: {image_path_for_gradio}")
+            # For Gradio type="messages", image content is just the path string
+            yield {"role": "assistant", "content": image_path_for_gradio} 
+            return 
+        except Exception as e:
+            print(f"DEBUG Gradio: Error saving PIL image from final_answer_content: {e}")
+            yield {"role": "assistant", "content": f"**Final Answer (Error displaying image):** {e}"}
+            return
 
-    # After the loop, step_log holds the final answer or the last step's log
-    final_answer_content = step_log 
     final_answer_processed = handle_agent_output_types(final_answer_content)
 
     if isinstance(final_answer_processed, AgentText):
-        yield gr.ChatMessage(role="assistant", content=f"**Final answer:**\n{final_answer_processed.to_string()}\n")
+        yield {"role": "assistant", "content": f"**Final Answer:**\n{final_answer_processed.to_string()}"}
     elif isinstance(final_answer_processed, AgentImage):
-        yield gr.ChatMessage(role="assistant", content={"path": final_answer_processed.to_string(), "mime_type": "image/png"})
+        image_path = final_answer_processed.to_string()
+        print(f"DEBUG Gradio (stream_to_gradio): AgentImage path: {image_path}")
+        if image_path and os.path.exists(image_path):
+             yield {"role": "assistant", "content": image_path} 
+        else:
+            err_msg = f"Error: Image path from AgentImage not found or invalid ('{image_path}')"
+            print(f"DEBUG Gradio: {err_msg}")
+            yield {"role": "assistant", "content": f"**Final Answer ({err_msg})**"}
     elif isinstance(final_answer_processed, AgentAudio):
-        yield gr.ChatMessage(role="assistant", content={"path": final_answer_processed.to_string(), "mime_type": "audio/wav"})
-    else:
-        yield gr.ChatMessage(role="assistant", content=f"**Final answer:** {str(final_answer_processed)}")
+        audio_path = final_answer_processed.to_string()
+        print(f"DEBUG Gradio (stream_to_gradio): AgentAudio path: {audio_path}")
+        if audio_path and os.path.exists(audio_path):
+            yield {"role": "assistant", "content": audio_path}
+        else:
+            err_msg = f"Error: Audio path from AgentAudio not found ('{audio_path}')"
+            print(f"DEBUG Gradio: {err_msg}")
+            yield {"role": "assistant", "content": f"**Final Answer ({err_msg})**"}
+    else: 
+        yield {"role": "assistant", "content": f"**Final Answer:**\n{str(final_answer_processed)}"}
 
 
 class GradioUI:
@@ -171,206 +168,199 @@ class GradioUI:
 
     def __init__(self, agent: MultiStepAgent, file_upload_folder: str | None = None):
         if not _is_package_available("gradio"):
-            raise ModuleNotFoundError(
-                "Please install 'gradio' extra to use the GradioUI: `pip install 'smolagents[gradio]'`"
-            )
+            raise ModuleNotFoundError("Install 'gradio': `pip install 'smolagents[gradio]'`")
         self.agent = agent
         self.file_upload_folder = file_upload_folder
         if self.file_upload_folder is not None:
-            if not os.path.exists(file_upload_folder):
-                os.makedirs(self.file_upload_folder, exist_ok=True) # Use makedirs
-        
-        self._latest_file_path_for_download = None # For download button state
+            if not os.path.exists(self.file_upload_folder):
+                os.makedirs(self.file_upload_folder, exist_ok=True)
+        self._latest_file_path_for_download = None
 
     def _check_for_created_file(self):
-        """Helper function to check interaction logs for a created file path."""
-        self._latest_file_path_for_download = None # Reset
+        self._latest_file_path_for_download = None 
         if hasattr(self.agent, 'interaction_logs') and self.agent.interaction_logs:
-            print(f"DEBUG UI: Checking {len(self.agent.interaction_logs)} interaction log entries.")
-            for log_entry in self.agent.interaction_logs:
-                if log_entry.get("tool_name") == "create_document":
-                    tool_output_value = log_entry.get("tool_output")
-                    print(f"DEBUG UI: Log for 'create_document', output: {tool_output_value}")
-                    if tool_output_value and isinstance(tool_output_value, str):
-                        if not tool_output_value.strip().startswith("ERROR:"):
-                            normalized_path = os.path.normpath(tool_output_value)
-                            if os.path.exists(normalized_path):
-                                self._latest_file_path_for_download = normalized_path
-                                print(f"DEBUG UI: File path for download set: {self._latest_file_path_for_download}")
-                                return True # Found a valid file
-                            else:
-                                print(f"DEBUG UI: Path from log ('{normalized_path}') does not exist.")
-                        else:
-                            print(f"DEBUG UI: 'create_document' tool reported error: {tool_output_value}")
+            print(f"DEBUG Gradio UI: Checking {len(self.agent.interaction_logs)} interaction log entries for created files.")
+            for log_entry in reversed(self.agent.interaction_logs): # Check recent logs first
+                if isinstance(log_entry, ActionStep) and hasattr(log_entry, 'tool_calls') and log_entry.tool_calls:
+                    for tool_call in log_entry.tool_calls:
+                        if tool_call.name == "create_document":
+                            tool_output_value = getattr(log_entry, 'observations', None)
+                            print(f"DEBUG Gradio UI: Log for 'create_document' call, observed output: {tool_output_value}")
+                            if tool_output_value and isinstance(tool_output_value, str):
+                                # Try to extract path if it's wrapped, e.g. by "Execution logs:"
+                                cleaned_output = re.sub(r"^Execution logs:\s*", "", tool_output_value).strip()
+                                path_match = re.search(r"(/tmp/[a-zA-Z0-9_]+/generated_document\.(?:docx|pdf|txt))", cleaned_output)
+                                extracted_path = path_match.group(1) if path_match else cleaned_output
+
+                                if not extracted_path.lower().startswith("error:"):
+                                    normalized_path = os.path.normpath(extracted_path)
+                                    if os.path.exists(normalized_path):
+                                        self._latest_file_path_for_download = normalized_path
+                                        print(f"DEBUG Gradio UI: File path for download set: {self._latest_file_path_for_download}")
+                                        return True 
+                                    else:
+                                        print(f"DEBUG Gradio UI: Path from 'create_document' log ('{normalized_path}') does not exist.")
+                                else:
+                                    print(f"DEBUG Gradio UI: 'create_document' tool reported error in observations: {extracted_path}")
+        print("DEBUG Gradio UI: No valid 'create_document' output found for download.")
         return False
 
+    def interact_with_agent(self, prompt_text: str, current_chat_tuples: list):
+        # current_chat_tuples is the history from the chatbot (list of lists/tuples)
+        # Convert to 'messages' format if needed, or adapt stream_to_gradio if chatbot is not type="messages"
+        # For type="messages", current_chat_tuples is already list of dicts.
+        
+        print(f"DEBUG Gradio: interact_with_agent called with prompt: '{prompt_text}'")
+        print(f"DEBUG Gradio: Current chat history (input): {current_chat_tuples}")
 
-    def interact_with_agent(self, prompt, messages_history, download_btn_state, file_output_state):
-        import gradio as gr
+        # Add user's new message to the chat history list
+        current_chat_messages = current_chat_tuples + [{"role": "user", "content": prompt_text}]
         
-        messages_history.append(gr.ChatMessage(role="user", content=prompt))
-        yield messages_history, gr.update(visible=False), gr.update(value=None, visible=False) # Hide download items initially
+        # Initial yield to show user message immediately and hide download items
+        yield current_chat_messages, gr.update(visible=False), gr.update(value=None, visible=False)
 
-        # Stream agent messages to chatbot
-        for msg in stream_to_gradio(self.agent, task=prompt, reset_agent_memory=False):
-            messages_history.append(msg)
-            yield messages_history, gr.update(visible=False), gr.update(value=None, visible=False) # Keep hidden during streaming
+        # Stream agent messages
+        agent_responses_for_history = []
+        for msg_dict in stream_to_gradio(self.agent, task=prompt_text, reset_agent_memory=False):
+            agent_responses_for_history.append(msg_dict)
+            # Yield progressively: current user message + all agent messages so far
+            yield current_chat_messages + agent_responses_for_history, gr.update(visible=False), gr.update(value=None, visible=False) 
 
         # After streaming all agent messages, check for created file
         file_found = self._check_for_created_file()
         
-        # Update UI based on whether a file was found
-        # Yielding final state for chatbot, download button, and file component
-        yield messages_history, gr.update(visible=file_found), gr.update(value=None, visible=False)
+        # Final state for UI components
+        final_chat_display = current_chat_messages + agent_responses_for_history
+        print(f"DEBUG Gradio: Final chat history for display: {final_chat_display}")
+        yield final_chat_display, gr.update(visible=file_found), gr.update(value=None, visible=False)
 
 
-    def upload_file(
-        self,
-        file,
-        file_uploads_log,
-        allowed_file_types=[
+    def upload_file(self, file, file_uploads_log_state):
+        if file is None: # No file selected
+            return gr.update(value="No file uploaded.", visible=True), file_uploads_log_state
+
+        # Ensure file_upload_folder exists (it should from __init__)
+        if not self.file_upload_folder or not os.path.exists(self.file_upload_folder):
+            os.makedirs(self.file_upload_folder, exist_ok=True) # Defensive check
+
+        allowed_file_types = [
             "application/pdf",
             "application/vnd.openxmlformats-officedocument.wordprocessingml.document",
             "text/plain",
-        ],
-    ):
-        import gradio as gr
-
-        if file is None:
-            return gr.Textbox("No file uploaded", visible=True), file_uploads_log
-
-        try:
-            mime_type, _ = mimetypes.guess_type(file.name)
-            if mime_type is None: # Fallback if guess_type returns None
-                mime_type = file.type # Gradio File object has a 'type' attribute
-        except Exception as e:
-            return gr.Textbox(f"Error: {e}", visible=True), file_uploads_log
+            "image/jpeg", "image/png", # Added image types
+        ]
         
+        # Gradio File object has 'name' (temp path) and 'orig_name'
+        original_name = file.orig_name if hasattr(file, 'orig_name') else os.path.basename(file.name)
+        
+        # Try to guess mime type from temp file name first, then from original name if needed
+        mime_type, _ = mimetypes.guess_type(file.name) 
+        if mime_type is None: # Fallback
+            mime_type, _ = mimetypes.guess_type(original_name)
+
         if mime_type not in allowed_file_types:
-            return gr.Textbox(f"File type '{mime_type}' disallowed", visible=True), file_uploads_log
+            return gr.update(value=f"File type '{mime_type or 'unknown'}' for '{original_name}' is disallowed.", visible=True), file_uploads_log_state
 
-        original_name = os.path.basename(file.name)
         sanitized_name = re.sub(r"[^\w\-.]", "_", original_name)
-        
-        # Ensure correct extension based on mime type, if possible
         base_name, current_ext = os.path.splitext(sanitized_name)
         
-        type_to_ext_map = {
+        type_to_ext_map = {v: k for k, v_list in mimetypes. প্রেফারেন্সেস.items() for v in v_list} # More robust ext map
+        type_to_ext_map.update({ # Manual overrides / common types
             "application/pdf": ".pdf",
             "application/vnd.openxmlformats-officedocument.wordprocessingml.document": ".docx",
-            "text/plain": ".txt",
-        }
+            "text/plain": ".txt", "image/jpeg": ".jpg", "image/png": ".png"
+        })
         expected_ext = type_to_ext_map.get(mime_type)
-        if expected_ext and current_ext.lower() != expected_ext:
+
+        if expected_ext and current_ext.lower() != expected_ext.lower():
             sanitized_name = base_name + expected_ext
         
-        file_path = os.path.join(self.file_upload_folder, sanitized_name)
-        shutil.copy(file.name, file_path) # file.name is the temp path of the uploaded file
-
-        return gr.Textbox(f"File uploaded: {file_path}", visible=True), file_uploads_log + [file_path]
-
-    def log_user_message(self, text_input, file_uploads_log):
-        # This function prepares the prompt that goes to the agent.
-        # It also clears the text_input box.
-        full_prompt = text_input
-        if file_uploads_log: # Check if list is not empty
-            full_prompt += (
-                f"\nYou have been provided with these files, which might be helpful or not: {file_uploads_log}"
-            )
-        return full_prompt, "" # Return the full prompt and an empty string to clear input
+        destination_path = os.path.join(self.file_upload_folder, sanitized_name)
+        
+        try:
+            shutil.copy(file.name, destination_path) # file.name is the temp path from Gradio
+            print(f"DEBUG Gradio: File '{original_name}' copied to '{destination_path}'")
+            updated_log = file_uploads_log_state + [destination_path]
+            return gr.update(value=f"Uploaded: {original_name} (as {sanitized_name})", visible=True), updated_log
+        except Exception as e:
+            print(f"DEBUG Gradio: Error copying uploaded file: {e}")
+            return gr.update(value=f"Error uploading {original_name}: {e}", visible=True), file_uploads_log_state
+
+
+    def log_user_message(self, text_input_value: str, current_file_uploads: list):
+        full_prompt = text_input_value
+        if current_file_uploads:
+            files_str = ", ".join([os.path.basename(f) for f in current_file_uploads])
+            full_prompt += f"\n\n[Uploaded files for context: {files_str}]"
+        print(f"DEBUG Gradio: Prepared prompt for agent: {full_prompt}")
+        return full_prompt, "" # Clears the text input box
 
     def prepare_and_show_download_file(self):
-        import gradio as gr
         if self._latest_file_path_for_download and os.path.exists(self._latest_file_path_for_download):
-            print(f"DEBUG UI: Preparing download for UI: {self._latest_file_path_for_download}")
+            print(f"DEBUG Gradio UI: Preparing download for UI component: {self._latest_file_path_for_download}")
             return gr.File.update(value=self._latest_file_path_for_download, 
                                   label=os.path.basename(self._latest_file_path_for_download), 
                                   visible=True)
         else:
-            print("DEBUG UI: No valid file path to prepare for download component.")
+            print("DEBUG Gradio UI: No valid file path to prepare for download component.")
             gr.Warning("No file available for download or path is invalid.")
             return gr.File.update(visible=False)
 
     def launch(self, **kwargs):
-        import gradio as gr
-
-        with gr.Blocks(fill_height=True, theme=gr.themes.Soft()) as demo: # Added a theme
-            # --- State Variables ---
-            # stored_messages is used to build the prompt for the agent, not directly for chatbot display here.
-            # The chatbot takes messages directly from interact_with_agent.
-            # We'll use chat_history_state for the chatbot's message list.
-            chat_history_state = gr.State([]) 
-            file_uploads_log = gr.State([]) # Tracks paths of uploaded files
+        with gr.Blocks(fill_height=True, theme=gr.themes.Soft(primary_hue=gr.themes.colors.blue)) as demo:
+            file_uploads_log_state = gr.State([]) 
+            prepared_prompt_for_agent = gr.State("")
 
-            # --- UI Layout ---
-            gr.Markdown("# Smol Talk with your Agent") # Title
+            gr.Markdown("#  agente inteligente")
 
             with gr.Row():
-                with gr.Column(scale=3): # Main chat area
-                    chatbot = gr.Chatbot(
+                with gr.Column(scale=3):
+                    chatbot_display = gr.Chatbot(
                         label="Agent Interaction",
-                        type="messages",
-                        # Bubble full width can make text hard to read, try default
-                        # bubble_full_width=False, 
-                        avatar_images=(
-                            None, # User avatar
-                            "https://huggingface.co/datasets/huggingface/brand-assets/resolve/main/hf-logo-round.png" # Agent avatar
-                        ),
-                        height=600 
+                        type="messages", 
+                        avatar_images=(None, "https://huggingface.co/datasets/huggingface/brand-assets/resolve/main/hf-logo-round.png"),
+                        height=600,
+                        show_copy_button=True,
+                        bubble_full_width=False
                     )
-                    text_input = gr.Textbox(
+                    text_message_input = gr.Textbox(
                         lines=1, 
                         label="Your Message to the Agent", 
-                        placeholder="Type your message and press Enter..."
+                        placeholder="Type your message and press Enter, or Shift+Enter for new line..."
                     )
                 
-                with gr.Column(scale=1): # Sidebar for uploads and downloads
+                with gr.Column(scale=1):
                     if self.file_upload_folder is not None:
                         gr.Markdown("### File Upload")
-                        upload_file_component = gr.File(label="Upload a supporting file")
-                        upload_status_display = gr.Textbox(label="Upload Status", interactive=False, visible=True, lines=2) # Make visible by default
-                        upload_file_component.upload( # Use 'upload' event for gr.File
+                        file_uploader = gr.File(label="Upload a supporting file (PDF, DOCX, TXT, JPG, PNG)")
+                        upload_status_text = gr.Textbox(label="Upload Status", interactive=False, lines=2, max_lines=4)
+                        file_uploader.upload(
                             self.upload_file,
-                            [upload_file_component, file_uploads_log],
-                            [upload_status_display, file_uploads_log],
+                            [file_uploader, file_uploads_log_state],
+                            [upload_status_text, file_uploads_log_state],
                         )
                     
                     gr.Markdown("### Generated File")
-                    # This button becomes visible if a file is created by the agent
-                    download_btn = gr.Button("Download Generated File", visible=False) 
-                    # This gr.File component becomes visible and populated when the button above is clicked
-                    file_output_display = gr.File(label="Downloadable Document", visible=False, interactive=False) 
-
-            # --- Event Handling ---
-            
-            # When user submits text_input:
-            # 1. log_user_message: prepares the prompt (text + file info), clears text_input.
-            #    The output 'prepared_prompt' is then passed to interact_with_agent.
-            # 2. interact_with_agent: streams agent's responses to chatbot, updates download button.
-            
-            # We need a state to hold the prepared prompt temporarily if log_user_message is separate
-            prepared_prompt_state = gr.State("")
-
-            text_input.submit(
-                self.log_user_message,
-                [text_input, file_uploads_log],
-                [prepared_prompt_state, text_input] # prepared_prompt_state gets the full prompt, text_input is cleared
+                    download_action_button = gr.Button("Download Generated File", visible=False) 
+                    file_download_display_component = gr.File(label="Downloadable Document", visible=False, interactive=False) 
+
+            # Event Handling Chain for Text Submission
+            text_message_input.submit(
+                self.log_user_message, # Step 1: Prepare prompt, clear input
+                [text_message_input, file_uploads_log_state],
+                [prepared_prompt_for_agent, text_message_input] 
             ).then(
-                self.interact_with_agent,
-                [prepared_prompt_state, chat_history_state, download_btn, file_output_display], # Pass current UI states
-                [chat_history_state, download_btn, file_output_display] # Update these UI states
+                self.interact_with_agent, # Step 2: Run agent, stream to chatbot, update download button
+                [prepared_prompt_for_agent, chatbot_display], 
+                [chatbot_display, download_action_button, file_download_display_component] 
             )
 
-            # When download_btn is clicked:
-            download_btn.click(
+            download_action_button.click(
                 self.prepare_and_show_download_file, 
-                [], # No inputs needed from UI for this action
-                [file_output_display] # Update the file_output_display component
+                [], 
+                [file_download_display_component] 
             )
 
-        # Launch the Gradio app
-        # Set share=False if running locally or on Spaces where share=True might be an issue
         demo.launch(debug=True, share=kwargs.get("share", False), **kwargs)
 
-
 __all__ = ["stream_to_gradio", "GradioUI"]