Update app.py

app.py

@@ -1,3 +1,7 @@
+
+
+
+
 import gradio as gr
 import weaviate
 from weaviate.embedded import EmbeddedOptions
@@ -5,9 +9,6 @@ import os
 from openai import OpenAI
 from dotenv import load_dotenv
 import textwrap
-from functools import lru_cache
-import asyncio
-import aiohttp
 
 # Load environment variables
 load_dotenv()
@@ -27,20 +28,19 @@ client = weaviate.Client(
 # Get the collection name from environment variable
 COLLECTION_NAME = os.getenv('WEAVIATE_COLLECTION_NAME')
 
-@lru_cache(maxsize=1000)
 def get_embedding(text):
     response = openai_client.embeddings.create(
         input=text,
-        model="text-embedding-3-small"
+        model="text-embedding-3-large"
     )
     return response.data[0].embedding
 
-async def search_multimodal(query: str, limit: int = 30, alpha: float = 0.6):
+def search_multimodal(query: str, limit: int = 30, alpha: float = 0.6):
     query_vector = get_embedding(query)
     
     try:
-        response = await asyncio.to_thread(
-            lambda: client.query
+        response = (
+            client.query
             .get(COLLECTION_NAME, ["content_type", "url", "source_document", "page_number",
                                    "paragraph_number", "text", "image_path", "description", "table_content"])
             .with_hybrid(
@@ -56,69 +56,70 @@ async def search_multimodal(query: str, limit: int = 30, alpha: float = 0.6):
         print(f"An error occurred during the search: {str(e)}")
         return []
 
-async def generate_response(query: str, context: str) -> str:
+def generate_response(query: str, context: str) -> str:
     prompt = f"""
 You are an AI assistant with extensive expertise in the semiconductor industry. Your knowledge spans a wide range of companies, technologies, and products, including but not limited to: System-on-Chip (SoC) designs, Field-Programmable Gate Arrays (FPGAs), Microcontrollers, Integrated Circuits (ICs), semiconductor manufacturing processes, and emerging technologies like quantum computing and neuromorphic chips.
 Use the following context, your vast knowledge, and the user's question to generate an accurate, comprehensive, and insightful answer. While formulating your response, follow these steps internally:
+
 Analyze the question to identify the main topic and specific information requested.
 Evaluate the provided context and identify relevant information.
 Retrieve additional relevant knowledge from your semiconductor industry expertise.
 Reason and formulate a response by combining context and knowledge.
 Generate a detailed response that covers all aspects of the query.
 Review and refine your answer for coherence and accuracy.
+
 In your output, provide only the final, polished response. Do not include your step-by-step reasoning or mention the process you followed.
 IMPORTANT: Ensure your response is grounded in factual information. Do not hallucinate or invent information. If you're unsure about any aspect of the answer or if the necessary information is not available in the provided context or your knowledge base, clearly state this uncertainty. It's better to admit lack of information than to provide inaccurate details.
 Your response should be:
+
 Thorough and directly address all aspects of the user's question
 Based solely on factual information from the provided context and your reliable knowledge
 Include specific examples, data points, or case studies only when you're certain of their accuracy
 Explain technical concepts clearly, considering the user may have varying levels of expertise
 Clearly indicate any areas where information is limited or uncertain
+
 Context: {context}
 User Question: {query}
 Based on the above context and your extensive knowledge of the semiconductor industry, provide your detailed, accurate, and grounded response below. Remember, only include information you're confident is correct, and clearly state any uncertainties: 
     """
 
-    async with aiohttp.ClientSession() as session:
-        async with session.post(
-            "https://api.openai.com/v1/chat/completions",
-            headers={
-                "Authorization": f"Bearer {os.getenv('OPENAI_API_KEY')}",
-                "Content-Type": "application/json"
-            },
-            json={
-                "model": "gpt-4-turbo-preview",
-                "messages": [
-                    {"role": "system", "content": "You are an expert Semi Conductor industry analyst"},
-                    {"role": "user", "content": prompt}
-                ],
-                "temperature": 0
-            }
-        ) as response:
-            result = await response.json()
-            return result['choices'][0]['message']['content']
-
-async def esg_analysis(user_query: str):
-    search_results = await search_multimodal(user_query)
-    
-    context = "\n\n".join(
-        f"{item['content_type'].capitalize()} from {item['source_document']} "
-        f"(Page {item['page_number']}): {item.get('text') or item.get('description')}"
-        for item in search_results
+    response = openai_client.chat.completions.create(
+        model="gpt-4o",
+        messages=[
+            {"role": "system", "content": "You are an expert Semi Conductor industry analyst"},
+            {"role": "user", "content": prompt}
+        ],
+        temperature=0
     )
 
-    response = await generate_response(user_query, context)
+    return response.choices[0].message.content
 
-    sources = [
-        {
+def esg_analysis(user_query: str):
+    search_results = search_multimodal(user_query)
+    
+    context = ""
+    for item in search_results:
+        if item['content_type'] == 'text':
+            context += f"Text from {item['source_document']} (Page {item['page_number']}, Paragraph {item['paragraph_number']}): {item['text']}\n\n"
+        elif item['content_type'] == 'image':
+            context += f"Image Description from {item['source_document']} (Page {item['page_number']}, Path: {item['image_path']}): {item['description']}\n\n"
+        elif item['content_type'] == 'table':
+            context += f"Table Description from {item['source_document']} (Page {item['page_number']}): {item['description']}\n\n"
+
+    response = generate_response(user_query, context)
+
+    sources = []
+    for item in search_results[:5]:  # Limit to top 5 sources
+        source = {
             "type": item.get("content_type", "Unknown"),
             "document": item.get("source_document", "N/A"),
             "page": item.get("page_number", "N/A"),
-            "paragraph": item.get("paragraph_number", "N/A") if item.get("content_type") == 'text' else None,
-            "image_path": item.get("image_path", "N/A") if item.get("content_type") == 'image' else None
         }
-        for item in search_results[:5]
-    ]
+        if item.get("content_type") == 'text':
+            source["paragraph"] = item.get("paragraph_number", "N/A")
+        elif item.get("content_type") == 'image':
+            source["image_path"] = item.get("image_path", "N/A")
+        sources.append(source)
 
     return response, sources
 
@@ -126,14 +127,27 @@ def wrap_text(text, width=120):
     return textwrap.fill(text, width=width)
 
 def gradio_interface(user_question):
-    ai_response, sources = asyncio.run(esg_analysis(user_question))
+    ai_response, sources = esg_analysis(user_question)
     
-    formatted_response = f"## AI Response\n{ai_response}"
+    # Format AI response
+    formatted_response = f"""
+## AI Response
+
+{ai_response}
+    """
     
-    source_text = "## Top 5 Sources\n\n" + "\n\n".join(
-        f"### Source {i+1}\n" + "\n".join(f"- **{k.title()}:** {v}" for k, v in source.items() if v is not None)
-        for i, source in enumerate(sources)
-    )
+    # Format sources
+    source_text = "## Top 5 Sources\n\n"
+    for i, source in enumerate(sources, 1):
+        source_text += f"### Source {i}\n"
+        source_text += f"- **Type:** {source['type']}\n"
+        source_text += f"- **Document:** {source['document']}\n"
+        source_text += f"- **Page:** {source['page']}\n"
+        if 'paragraph' in source:
+            source_text += f"- **Paragraph:** {source['paragraph']}\n"
+        if 'image_path' in source:
+            source_text += f"- **Image Path:** {source['image_path']}\n"
+        source_text += "\n"
     
     return formatted_response, source_text
 
@@ -146,7 +160,7 @@ iface = gr.Interface(
     ],
     title="Semiconductor Industry ESG Analysis",
     description="Ask questions about the semiconductor industry and get AI-powered answers with sources.",
-    flagging_dir="/app/flagged"
+    flagging_dir="/app/flagged"  # Specify the flagging directory
 )
 
 if __name__ == "__main__":