Update app.py

app.py

@@ -1,148 +1,63 @@
+import gradio as gr
+import httpx
+import time
 from datetime import datetime
 from llama_index.core import VectorStoreIndex, SimpleDirectoryReader
-from llama_index.embeddings.huggingface import HuggingFaceEmbedding
 from llama_parse import LlamaParse
+from llama_index.embeddings.huggingface import HuggingFaceEmbedding
 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
-load_dotenv()
-
-llm_models = [
-    "tiiuae/falcon-7b-instruct",
-]
-
-embed_models = [
-    "BAAI/bge-small-en-v1.5",  # 33.4M
-]
-
-# Global variable for selected model
-selected_llm_model_name = llm_models[0]  # Default to the first model in the list
-selected_embed_model_name = embed_models[0] # Default to the first model in the list
-vector_index = None
-
-# Initialize the parser
-parser = LlamaParse(api_key="llx-zKtsC5UBLs8DOApOsLluXMBdQhC75ea0Vs80SmPSjsmDzuhh", result_type='markdown')
-# Define file extractor with various common extensions
-file_extractor = {
-    '.pdf': parser,  # PDF documents
-    '.docx': parser,  # Microsoft Word documents
-    '.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)
-  
-
-    # Image files for OCR processing
-    '.jpg': parser,  # JPEG images
-    '.jpeg': parser,  # JPEG images
-    '.png': parser,  # PNG images
-  
-
-    # Scanned documents in image formats
-    '.webp': parser,  # WebP images
-    '.svg': parser,  # SVG files (vector format, may contain embedded text)
-}
-
-
-# File processing function
-def load_files(file_path: str, embed_model_name: str):
-    try:
-        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}"
-    except Exception as e:
-        return f"An error occurred: {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}"
-
-
-# Respond function that uses the globally set selected model
-def respond(message, history):
-    try:
-        # 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
-            token=os.getenv("TOKEN")
-        )
-
-        # Set up the query engine with the selected LLM
-        query_engine = vector_index.as_query_engine(llm=llm)
-        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)}"
-    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}"
-
-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")
-
-# UI Setup
-with gr.Blocks(theme=gr.themes.Soft(font=[gr.themes.GoogleFont("Roboto Mono")]), css='footer {visibility: hidden}') as demo:
-    gr.Markdown("# DocBot📄🤖")
-    with gr.Tabs():
-        with gr.TabItem("Intro"):
-            gr.Markdown(md.description)
-
-        with gr.TabItem("DocBot"):
-            with gr.Accordion("=== IMPORTANT: READ ME FIRST ===", open=False):
-                guid = gr.Markdown(md.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)
-                    )
-    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])
 
-# Launch the demo with a public link option
+# LLM ve Parser Başlatma
+llm = HuggingFaceInferenceAPI(model_name="meta-llama/Llama-3.2-1B")
+parser = LlamaParse(api_key='llx-zKtsC5UBLs8DOApOsLluXMBdQhC75ea0Vs80SmPSjsmDzuhh', result_type='markdown')
+
+# PDF dosyasını yükleyip indexleme işlemi
+# Bu kısımda pdf zaten sistemde yüklü olduğundan 'data/' dizininde olduğundan emin olun
+file_extractor = {'.pdf': parser}
+documents = SimpleDirectoryReader('data/', file_extractor=file_extractor).load_data()
+
+# Embedding Modeli ve Query Engine Başlatma
+embed_model = HuggingFaceEmbedding(model_name="BAAI/bge-small-en-v1.5")
+vector_index = VectorStoreIndex.from_documents(documents, embed_model=embed_model)
+query_engine = vector_index.as_query_engine(llm=llm)
+
+# Sorgu işlemi için retry mekanizması
+def query_with_retry(query, max_retries=3, wait_time=5):
+    for attempt in range(max_retries):
+        try:
+            start_time = datetime.now()
+            response = query_engine.query(query)
+            end_time = datetime.now()
+            duration = (end_time - start_time).total_seconds()
+            print(f"Query completed in {duration:.2f} seconds.\n {response}")
+            return response
+        except httpx.ReadTimeout:
+            if attempt < max_retries - 1:
+                print(f"Timeout occurred. Retrying in {wait_time} seconds...")
+                time.sleep(wait_time)
+            else:
+                raise
+        except Exception as e:
+            print(f"An error occurred: {e}")
+            break
+
+# Gradio arayüzü
+def respond_to_pdf(query):
+    return query_with_retry(query)
+
+# Basit Gradio Arayüzü
+with gr.Blocks() as demo:
+    gr.Markdown("# PDF DocBot")
+    
+    # Kullanıcıların PDF'e sorular sorması için UI öğeleri
+    query_input = gr.Textbox(label="Step-1: Ask a question about the pre-uploaded PDF")
+    output = gr.Textbox(label="Response")
+
+    btn = gr.Button("Submit")
+    
+    # Soru sorulduğunda modelin cevabını almak için
+    btn.click(fn=respond_to_pdf, inputs=[query_input], outputs=output)
+
+# Uygulamayı başlat
 if __name__ == "__main__":
-    demo.launch(share=True)
+    demo.launch()