Update app.py

app.py

@@ -6,7 +6,6 @@ from llama_index.llms.huggingface_api import HuggingFaceInferenceAPI
 import os
 from dotenv import load_dotenv
 import gradio as gr
-import markdowm as md
 import base64
 
 # Load environment variables
@@ -17,32 +16,6 @@ llm_models = [
     "meta-llama/Meta-Llama-3-8B-Instruct",
     "mistralai/Mistral-7B-Instruct-v0.2",
     "tiiuae/falcon-7b-instruct",
-    # "deepseek-ai/DeepSeek-R1-Distill-Qwen-32B",
-    # "deepseek-ai/deepseek-vl2",  ## 54GB > 10GB
-    # "deepseek-ai/deepseek-vl2-small",  ## 32GB > 10GB
-    # "deepseek-ai/deepseek-vl2-tiny",  ## high response time
-    # "deepseek-ai/deepseek-llm-7b-chat",  ## 13GB > 10GB
-    # "deepseek-ai/deepseek-math-7b-instruct",  ## 13GB > 10GB
-    # "deepseek-ai/deepseek-coder-33b-instruct",  ## 66GB > 10GB
-    # "deepseek-ai/DeepSeek-R1-Zero",  ## 688GB > 10GB
-    # "mistralai/Mixtral-8x22B-Instruct-v0.1",    ## 281GB>10GB
-    # "NousResearch/Yarn-Mistral-7b-64k",  ## 14GB>10GB
-    # "impira/layoutlm-document-qa",  ## ERR
-    # "Qwen/Qwen1.5-7B",      ## 15GB
-    # "Qwen/Qwen2.5-3B",      ## high response time
-    # "google/gemma-2-2b-jpn-it",   ## high response time
-    # "impira/layoutlm-invoices",   ## bad req
-    # "google/pix2struct-docvqa-large",  ## bad req
-    # "google/gemma-7b-it", ## 17GB > 10GB
-    # "google/gemma-2b-it",  ## high response time
-    # "HuggingFaceH4/zephyr-7b-beta",   ## high response time
-    # "HuggingFaceH4/zephyr-7b-gemma-v0.1",     ## bad req
-    # "microsoft/phi-2",    ## high response time
-    # "TinyLlama/TinyLlama-1.1B-Chat-v1.0",     ## high response time
-    # "mosaicml/mpt-7b-instruct",     ## 13GB>10GB
-    # "google/flan-t5-xxl" ## high respons time
-    # "NousResearch/Yarn-Mistral-7b-128k",  ## 14GB>10GB
-    # "Qwen/Qwen2.5-7B-Instruct",     ## 15GB>10GB
 ]
 
 embed_models = [
@@ -59,6 +32,7 @@ vector_index = None
 
 # Initialize the parser
 parser = LlamaParse(api_key=os.getenv("LLAMA_INDEX_API"), result_type='markdown')
+
 # Define file extractor with various common extensions
 file_extractor = {
     '.pdf': parser,  # PDF documents
@@ -66,62 +40,109 @@ file_extractor = {
     '.doc': parser,  # Older Microsoft Word documents
     '.txt': parser,  # Plain text files
     '.csv': parser,  # Comma-separated values files
-    '.xlsx': parser,  # Microsoft Excel files (requires additional processing for tables)
-    '.pptx': parser,  # Microsoft PowerPoint files (for slides)
-    '.html': parser,  # HTML files (web pages)
-    # '.rtf': parser,  # Rich Text Format files
-    # '.odt': parser,  # OpenDocument Text files
-    # '.epub': parser,  # ePub files (e-books)
-
-    # Image files for OCR processing
+    '.xlsx': parser,  # Microsoft Excel files
+    '.pptx': parser,  # Microsoft PowerPoint files
+    '.html': parser,  # HTML files
     '.jpg': parser,  # JPEG images
     '.jpeg': parser,  # JPEG images
     '.png': parser,  # PNG images
-    # '.bmp': parser,  # Bitmap images
-    # '.tiff': parser,  # TIFF images
-    # '.tif': parser,  # TIFF images (alternative extension)
-    # '.gif': parser,  # GIF images (can contain text)
-
-    # Scanned documents in image formats
     '.webp': parser,  # WebP images
-    '.svg': parser,  # SVG files (vector format, may contain embedded text)
+    '.svg': parser,  # SVG files
 }
 
+# Markdown content definitions
+description = """
+## Welcome to DocBot 📄🤖
+
+DocBot is an intelligent document analysis tool that can help you extract insights from various document formats including:
+- PDF documents
+- Word documents (.docx, .doc)
+- Text files
+- CSV files
+- Excel files
+- PowerPoint presentations
+- HTML files
+- Images with text (JPG, PNG, WebP, SVG)
+
+Simply upload your document, select your preferred embedding model and LLM, then start asking questions!
+"""
+
+guide = """
+### How to Use DocBot:
+
+1. **Upload Document**: Choose any supported file format
+2. **Select Embedding Model**: Choose from available embedding models (BAAI/bge-small-en-v1.5 is recommended for most cases)
+3. **Submit**: Click submit to process your document
+4. **Select LLM**: Choose your preferred language model
+5. **Ask Questions**: Start chatting with your document!
+
+### Tips:
+- Smaller embedding models (like bge-small-en-v1.5) are faster but may be less accurate
+- Larger models provide better understanding but take more time
+- Be specific in your questions for better results
+"""
+
+footer = """
+<div style="text-align: center; margin-top: 20px; padding: 20px; border-top: 1px solid #ddd;">
+    <p>Built with ❤️ using LlamaIndex and Gradio</p>
+    <div style="display: flex; justify-content: center; gap: 20px; margin-top: 10px;">
+        <a href="https://github.com" target="_blank">
+            <img src="data:image/png;base64,{0}" alt="GitHub" style="width: 24px; height: 24px;">
+        </a>
+        <a href="https://linkedin.com" target="_blank">
+            <img src="data:image/png;base64,{1}" alt="LinkedIn" style="width: 24px; height: 24px;">
+        </a>
+        <a href="https://your-website.com" target="_blank">
+            <img src="data:image/png;base64,{2}" alt="Website" style="width: 24px; height: 24px;">
+        </a>
+    </div>
+</div>
+"""
 
 # File processing function
 def load_files(file_path: str, embed_model_name: str):
     try:
+        if not file_path:
+            return "Please select a file first."
+        
+        if not embed_model_name:
+            return "Please select an embedding model."
+            
         global vector_index
         document = SimpleDirectoryReader(input_files=[file_path], file_extractor=file_extractor).load_data()
         embed_model = HuggingFaceEmbedding(model_name=embed_model_name)
         vector_index = VectorStoreIndex.from_documents(document, embed_model=embed_model)
         print(f"Parsing done for {file_path}")
         filename = os.path.basename(file_path)
-        return f"Ready to give response on {filename}"
+        return f"✅ Ready to answer questions about: {filename}"
     except Exception as e:
-        return f"An error occurred: {e}"
-
+        return f"❌ An error occurred: {str(e)}"
 
 # Function to handle the selected model from dropdown
 def set_llm_model(selected_model):
     global selected_llm_model_name
-    selected_llm_model_name = selected_model  # Update the global variable
-    # print(f"Model selected: {selected_model_name}")
-    # return f"Model set to: {selected_model_name}"
-
+    if selected_model:
+        selected_llm_model_name = selected_model
+    return f"LLM set to: {selected_model}"
 
 # Respond function that uses the globally set selected model
 def respond(message, history):
     try:
+        if not vector_index:
+            return "Please upload and process a document first."
+            
+        if not message.strip():
+            return "Please enter a question."
+            
         # Initialize the LLM with the selected model
         llm = HuggingFaceInferenceAPI(
             model_name=selected_llm_model_name,
-            contextWindow=8192,  # Context window size (typically max length of the model)
-            maxTokens=1024,  # Tokens per response generation (512-1024 works well for detailed answers)
-            temperature=0.3,  # Lower temperature for more focused answers (0.2-0.4 for factual info)
-            topP=0.9,  # Top-p sampling to control diversity while retaining quality
-            frequencyPenalty=0.5,  # Slight penalty to avoid repetition
-            presencePenalty=0.5,  # Encourages exploration without digressing too much
+            contextWindow=8192,
+            maxTokens=1024,
+            temperature=0.3,
+            topP=0.9,
+            frequencyPenalty=0.5,
+            presencePenalty=0.5,
             token=os.getenv("TOKEN")
         )
 
@@ -130,58 +151,151 @@ def respond(message, history):
         bot_message = query_engine.query(message)
 
         print(f"\n{datetime.now()}:{selected_llm_model_name}:: {message} --> {str(bot_message)}\n")
-        return f"{selected_llm_model_name}:\n{str(bot_message)}"
+        return f"**{selected_llm_model_name}:**\n\n{str(bot_message)}"
     except Exception as e:
-        if str(e) == "'NoneType' object has no attribute 'as_query_engine'":
-            return "Please upload a file."
-        return f"An error occurred: {e}"
+        return f"❌ An error occurred: {str(e)}"
 
-def encode_image(image_path):
-    with open(image_path, "rb") as image_file:
-        return base64.b64encode(image_file.read()).decode('utf-8')
-
-# Encode the images
-github_logo_encoded = encode_image("Images/github-logo.png")
-linkedin_logo_encoded = encode_image("Images/linkedin-logo.png")
-website_logo_encoded = encode_image("Images/ai-logo.png")
+def encode_image_safe(image_path):
+    """Safely encode image, return empty string if file doesn't exist"""
+    try:
+        if os.path.exists(image_path):
+            with open(image_path, "rb") as image_file:
+                return base64.b64encode(image_file.read()).decode('utf-8')
+    except Exception:
+        pass
+    return ""
 
-# theme=gr.themes.Soft(font=[gr.themes.GoogleFont("Roboto Mono")]), 
+# Encode the images (with fallback for missing images)
+github_logo_encoded = encode_image_safe("Images/github-logo.png")
+linkedin_logo_encoded = encode_image_safe("Images/linkedin-logo.png")
+website_logo_encoded = encode_image_safe("Images/ai-logo.png")
 
 # UI Setup
-with gr.Blocks(css='footer {visibility: hidden}') as demo:
-    gr.Markdown("# DocBot📄🤖")
+with gr.Blocks(
+    theme=gr.themes.Soft(),
+    css='footer {visibility: hidden}',
+    title="DocBot - Document Analysis Assistant"
+) as demo:
+    
+    gr.Markdown("# DocBot 📄🤖")
+    gr.Markdown("*Intelligent Document Analysis Assistant*")
+    
     with gr.Tabs():
-        with gr.TabItem("Intro"):
-            gr.Markdown(md.description)
+        with gr.TabItem("📖 Introduction"):
+            gr.Markdown(description)
 
-        with gr.TabItem("DocBot"):
-            with gr.Accordion("=== IMPORTANT: READ ME FIRST ===", open=False):
-                guid = gr.Markdown(md.guide)
+        with gr.TabItem("🤖 DocBot"):
+            with gr.Accordion("📋 Quick Start Guide", open=False):
+                gr.Markdown(guide)
+                
             with gr.Row():
                 with gr.Column(scale=1):
-                    file_input = gr.File(file_count="single", type='filepath', label="Step-1: Upload document")
-                    # gr.Markdown("Dont know what to select check out in Intro tab")
-                    embed_model_dropdown = gr.Dropdown(embed_models, label="Step-2: Select Embedding", interactive=True)
-                    with gr.Row():
-                        btn = gr.Button("Submit", variant='primary')
-                        clear = gr.ClearButton()
-                    output = gr.Text(label='Vector Index')
-                    llm_model_dropdown = gr.Dropdown(llm_models, label="Step-3: Select LLM", interactive=True)
-                with gr.Column(scale=3):
-                    gr.ChatInterface(
-                        fn=respond,
-                        chatbot=gr.Chatbot(height=500),
-                        # theme = "soft",
-                        show_progress='full',
-                        # cache_mode='lazy',
-                        textbox=gr.Textbox(placeholder="Step-4: Ask me questions on the uploaded document!", container=False)
+                    with gr.Group():
+                        gr.Markdown("### Document Processing")
+                        file_input = gr.File(
+                            file_count="single", 
+                            type='filepath', 
+                            label="Step 1: Upload Document",
+                            file_types=['.pdf', '.docx', '.doc', '.txt', '.csv', '.xlsx', '.pptx', '.html', '.jpg', '.jpeg', '.png', '.webp', '.svg']
+                        )
+                        
+                        embed_model_dropdown = gr.Dropdown(
+                            choices=embed_models, 
+                            label="Step 2: Select Embedding Model", 
+                            interactive=True,
+                            value=embed_models[0]
+                        )
+                        
+                        with gr.Row():
+                            btn = gr.Button("🚀 Process Document", variant='primary', size="lg")
+                            clear = gr.ClearButton("🗑️ Clear", size="lg")
+                        
+                        output = gr.Textbox(
+                            label='Processing Status', 
+                            interactive=False,
+                            placeholder="Upload a document and click 'Process Document' to begin..."
+                        )
+                    
+                    with gr.Group():
+                        gr.Markdown("### Model Selection")
+                        llm_model_dropdown = gr.Dropdown(
+                            choices=llm_models, 
+                            label="Step 3: Select Language Model", 
+                            interactive=True,
+                            value=llm_models[0]
+                        )
+                        llm_status = gr.Textbox(
+                            label="Selected Model",
+                            interactive=False,
+                            value=f"LLM set to: {llm_models[0]}"
+                        )
+                
+                with gr.Column(scale=2):
+                    gr.Markdown("### Chat with Your Document")
+                    chatbot = gr.Chatbot(
+                        height=600,
+                        placeholder="Process a document first, then start asking questions!",
+                        show_label=False
+                    )
+                    
+                    msg = gr.Textbox(
+                        placeholder="Step 4: Ask questions about your document...",
+                        container=False,
+                        scale=7
                     )
-    gr.HTML(md.footer.format(github_logo_encoded, linkedin_logo_encoded, website_logo_encoded))
-    # Set up Gradio interactions
-    llm_model_dropdown.change(fn=set_llm_model, inputs=llm_model_dropdown)
-    btn.click(fn=load_files, inputs=[file_input, embed_model_dropdown], outputs=output)
-    clear.click(lambda: [None] * 3, outputs=[file_input, embed_model_dropdown, output])
+                    
+                    with gr.Row():
+                        submit_btn = gr.Button("Send", variant="primary")
+                        clear_chat = gr.ClearButton([msg, chatbot], value="Clear Chat")
+
+    # Add footer if images exist
+    if any([github_logo_encoded, linkedin_logo_encoded, website_logo_encoded]):
+        gr.HTML(footer.format(github_logo_encoded, linkedin_logo_encoded, website_logo_encoded))
+    
+    # Set up event handlers
+    def chat_respond(message, history):
+        if not message.strip():
+            return history, ""
+        
+        response = respond(message, history)
+        history.append([message, response])
+        return history, ""
+    
+    # Event bindings
+    llm_model_dropdown.change(
+        fn=set_llm_model, 
+        inputs=[llm_model_dropdown], 
+        outputs=[llm_status]
+    )
+    
+    btn.click(
+        fn=load_files, 
+        inputs=[file_input, embed_model_dropdown], 
+        outputs=[output]
+    )
+    
+    submit_btn.click(
+        fn=chat_respond,
+        inputs=[msg, chatbot],
+        outputs=[chatbot, msg]
+    )
+    
+    msg.submit(
+        fn=chat_respond,
+        inputs=[msg, chatbot], 
+        outputs=[chatbot, msg]
+    )
+    
+    clear.click(
+        lambda: [None, None, ""], 
+        outputs=[file_input, embed_model_dropdown, output]
+    )
 
-# Launch the demo with a public link option
+# Launch the demo
 if __name__ == "__main__":
-    demo.launch(share=True)
+    demo.launch(
+        share=True,
+        server_name="0.0.0.0",
+        server_port=7860,
+        show_error=True
+    )