Update app.py

app.py

@@ -1,6 +1,7 @@
+
+import os
 import gradio as gr
 from huggingface_hub import InferenceClient
-import os
 import requests
 import json
 import pytesseract
@@ -10,83 +11,142 @@ from io import BytesIO
 import docx
 import ntpath
 
-# Initialize clients
-API_KEY = os.environ.get("HF_API_KEY")
-client = InferenceClient(token=API_KEY)
+###############################################################################
+# REPLACE WITH YOUR DEFAULT (OPEN-SOURCE) DEEPSEEK API KEY
+###############################################################################
+OPEN_SOURCE_DEEPSEEK_KEY = "YOUR_DEFAULT_API_KEY_HERE"
 
-def process_file(filepath):
-    """Handle different file types and extract text."""
-    if not filepath:
-        return ""
-    
-    ext = os.path.splitext(filepath)[1].lower()
-    
-    try:
-        if ext in ['.png', '.jpg', '.jpeg']:
-            image = Image.open(filepath)
-            text = pytesseract.image_to_string(image)
-            return f"IMAGE CONTENT:\n{text}"
-        elif ext == '.pdf':
-            pdf_reader = PyPDF2.PdfReader(filepath)
-            text = "\n".join(page.extract_text() for page in pdf_reader.pages)
-            return f"PDF CONTENT:\n{text}"
-        elif ext == '.docx':
-            doc = docx.Document(filepath)
-            text = "\n".join(para.text for para in doc.paragraphs)
-            return f"DOCUMENT CONTENT:\n{text}"
-        else:
-            return "Unsupported file type"
-    except Exception as e:
-        print(f"File processing error: {e}")
-        return "Error reading file"
-
-def chat(message, history, filepath):
-    file_content = process_file(filepath) if filepath else ""
-    
-    full_prompt = f"""
-    {file_content}
-    
-    User Message: {message}
-    
-    Please respond considering both the message and any attached documents:
+def chat_with_deepseek(
+    user_message,    # The user’s message to the chatbot
+    history,         # Chat history (handled by Gradio’s ChatInterface)
+    system_message,  # The system message from the left panel
+    user_api_key,    # The user-provided optional HF API key
+    max_tokens,      # The “Max new tokens” slider
+    temperature,     # The “Temperature” slider
+    top_p            # The “Top-p” slider
+):
+    """
+    This function is called every time a user submits a new message.
+    It uses either the user’s HF API key if provided, or a default
+    open-source DeepSeek API key if empty.
     """
 
-    generate_kwargs = dict(
-        temperature=0.7,
-        max_new_tokens=2000,
-        top_p=0.95,
-        repetition_penalty=1.2,
+    # Decide which key to use
+    final_api_key = user_api_key.strip() if user_api_key else OPEN_SOURCE_DEEPSEEK_KEY
+
+    # Initialize InferenceClient with the chosen API key
+    client = InferenceClient(token=final_api_key)
+
+    # Build the prompt or system instruction
+    # You can format your prompt however you like; here’s a simple example:
+    prompt = (
+        f"{system_message.strip()}\n\n"    # System instructions at the top
+        f"User: {user_message}\n"
+        "Assistant:"
     )
-    
-    # Generate response
+
+    # Configure generation parameters
+    generation_params = dict(
+        temperature=temperature,
+        max_new_tokens=int(max_tokens),
+        top_p=top_p,
+        repetition_penalty=1.1,  # If you want to add a penalty
+    )
+
+    # Stream the response
+    partial_response = ""
     stream = client.text_generation(
-        full_prompt,
+        prompt,
         stream=True,
         details=True,
-        **generate_kwargs
+        **generation_params
     )
-    
-    partial_message = ""
-    for response in stream:
-        if response.token.special:
+    for chunk in stream:
+        if chunk.token.special:
             continue
-        partial_message += response.token.text
-        yield partial_message
+        partial_response += chunk.token.text
+        # Yield partial chunks so user sees a "streaming" effect
+        yield partial_response
+
+
+###############################################################################
+# BUILD THE GRADIO INTERFACE
+###############################################################################
 
 with gr.Blocks(theme="soft") as demo:
-    gr.Markdown("# DeepSeek-R1 Assistant with File Support")
-    gr.Markdown("Upload images, PDFs, or docs and chat about them!")
+    # Top Title or Branding
+    gr.Markdown(
+        """
+        <h1 style="text-align:center; margin-bottom: 5px;">
+            <b>balianone.com</b>
+        </h1>
+        <h3 style="text-align:center; margin-top: 0px;">
+            Chat with DeepSeek-R1
+        </h3>
+        """,
+        elem_id="title"
+    )
     
     with gr.Row():
-        file_input = gr.File(label="Upload File (PDF/Image/Doc)", type="filepath")
-        
-    chatbot = gr.ChatInterface(
-        fn=chat,
-        additional_inputs=[file_input],
-        type="messages",
-        examples=[]
-    )
+        # Left Column (System message, HF API key, sliders)
+        with gr.Column(scale=1, min_width=250):
+            system_message = gr.Textbox(
+                label="System message", 
+                value="You are a friendly Chatbot created by Bali"
+            )
+            user_api_key = gr.Textbox(
+                label="HF API Key (optional)", 
+                type="password",
+                placeholder="Leave blank to use open-source DeepSeek key"
+            )
+            
+            gr.Markdown(
+                "If error occurs, you can set your own API key. You’ll get R1 premium access."
+            )
+            
+            max_tokens = gr.Slider(
+                minimum=1, 
+                maximum=4000, 
+                step=1, 
+                value=4000, 
+                label="Max new tokens"
+            )
+            temperature = gr.Slider(
+                minimum=0.1, 
+                maximum=4.0, 
+                value=0.7, 
+                step=0.1, 
+                label="Temperature"
+            )
+            top_p = gr.Slider(
+                minimum=0.1,
+                maximum=1.0,
+                value=0.95,
+                step=0.01,
+                label="Top-p (nucleus sampling)"
+            )
+
+        # Right Column (Chat interface)
+        with gr.Column(scale=3):
+            chatbot = gr.ChatInterface(
+                fn=chat_with_deepseek, 
+                # Additional inputs that feed into your chat function:
+                additional_inputs=[system_message, user_api_key, max_tokens, temperature, top_p],
+                
+                # Use the newer messages format to avoid warnings
+                type="messages",
+                
+                # You can customize how the chat area looks here as well:
+                chatbot_base_url=None,  # Keep local
+                height=550,
+                title="DeepSeek-R1 Assistant"
+            )
+
+    # Launch the app
+    demo.launch()
+
+
+
 
-demo.launch()