import streamlit as st import google.generativeai as genai import time import PyPDF2 from datetime import datetime from io import BytesIO from knowledge_base import get_industrial_answer from pdf2image import convert_from_bytes # Configuration try: genai.configure( api_key=st.secrets["GOOGLE_API_KEY"], client_options={"api_endpoint": "generativelanguage.googleapis.com/v1beta"} ) model = genai.GenerativeModel( 'gemini-1.5-flash-latest', generation_config={ "temperature": 0.3, "max_output_tokens": 500, "response_mime_type": "text/plain" }, safety_settings={ "HARASSMENT": "BLOCK_NONE", "HATE": "BLOCK_NONE", "SEXUAL": "BLOCK_NONE", "DANGEROUS": "BLOCK_NONE" } ) except Exception as config_error: st.error(f"Configuration Error: {str(config_error)}") st.stop() # Custom CSS for dark/light mode def inject_custom_css(): st.markdown(f""" """, unsafe_allow_html=True) # System prompt template SYSTEM_PROMPT = """**Invoice Analysis Expert** Analyze the provided document and extract: 1. Invoice Number 2. Total Amount 3. Due Date 4. Vendor/Customer Name 5. Itemized Charges Follow these rules: - Present data in markdown table - Highlight important dates/amounts in bold - Cross-verify with company policies - Current Date: {date} - Language: {language} Document Content: {file_content} User Question: {question} """ def process_uploaded_files(files): """Process uploaded files for analysis""" processed = [] if not files: return processed for file in files: try: file_bytes = file['bytes'] mime_type = file['type'] if mime_type == "application/pdf": try: # Try native PDF handling processed.append( genai.upload_file( mime_type="application/pdf", file=BytesIO(file_bytes) ) except Exception as pdf_error: # Fallback to text extraction pdf_text = "" pdf_reader = PyPDF2.PdfReader(BytesIO(file_bytes)) for page in pdf_reader.pages[:3]: # First 3 pages only page_text = page.extract_text() if page_text: pdf_text += page_text + "\n" if pdf_text.strip(): processed.append(pdf_text) else: # Convert PDF to images images = convert_from_bytes(file_bytes, first_page=0, last_page=2) for img in images: img_byte_arr = BytesIO() img.save(img_byte_arr, format='JPEG') processed.append( genai.upload_file( mime_type="image/jpeg", file=img_byte_arr ) ) elif mime_type.startswith("image/"): processed.append( genai.upload_file( mime_type=mime_type, file=BytesIO(file_bytes) ) ) except Exception as e: st.error(f"Error processing {file['name']}: {str(e)}") continue return processed def generate_response(prompt: str, language: str, files: list) -> str: """Generate response with document analysis""" start_time = time.time() try: # Process files first file_contents = process_uploaded_files(files) if not file_contents: return "⚠️ Could not process any uploaded documents" # Build prompt formatted_prompt = SYSTEM_PROMPT.format( date=datetime.now().strftime("%Y-%m-%d"), language=language, file_content="\n\n".join([ f"Document {i+1} Content:\n" + (content if isinstance(content, str) else "[Image Content]") for i, content in enumerate(file_contents) ]), question=prompt ) # Generate response response = model.generate_content( [formatted_prompt] + file_contents, request_options={"timeout": 30} ) return response.text if response.text else "No relevant information found" except Exception as e: return f"Analysis Error: {str(e)}" finally: if time.time() - start_time > 25: st.error("Analysis timeout. Try smaller documents or specific questions.") # Session State Management if "messages" not in st.session_state: st.session_state.messages = [] if "files" not in st.session_state: st.session_state.files = [] if "dark_mode" not in st.session_state: st.session_state.dark_mode = False if "language" not in st.session_state: st.session_state.language = "English" # Theme Configuration theme = { "dark": { "bg": "#1a1a1a", "text": "#ffffff", "sidebar": "#2d2d2d", "msg": "#333333" }, "light": { "bg": "#ffffff", "text": "#000000", "sidebar": "#f0f0f0", "msg": "#f8f9fa" } } # Apply theme mode = "dark" if st.session_state.dark_mode else "light" st.session_state.bg_color = theme[mode]["bg"] st.session_state.text_color = theme[mode]["text"] st.session_state.sidebar_bg = theme[mode]["sidebar"] st.session_state.msg_bg = theme[mode]["msg"] # Sidebar Controls with st.sidebar: st.title("⚙️ Settings") # Theme Toggle st.session_state.dark_mode = st.toggle( "🌙 Dark Mode", value=st.session_state.dark_mode ) # Language Selector st.session_state.language = st.selectbox( "🌐 Select Language", ["English", "Spanish", "French", "German", "Chinese"], index=0 ) # Chat History st.subheader("📚 Chat History") history_expander = st.expander("View History") with history_expander: for idx, msg in enumerate(st.session_state.messages): if msg["role"] == "user": st.markdown(f"**You ({idx+1}):** {msg['content']}") else: st.markdown(f"**TARS ({idx+1}):** {msg['content']}") # File Uploader st.subheader("📁 Upload Documents") uploaded_files = st.file_uploader( "Select PDF/Image files", type=["pdf", "png", "jpg", "jpeg"], accept_multiple_files=True, label_visibility="collapsed" ) if uploaded_files: st.session_state.files = [{ "name": file.name, "type": file.type, "bytes": file.read() } for file in uploaded_files] # Main Interface st.title("🤖 Industrial Document Analyst") st.caption("Your AI-powered document analysis assistant") # Chat Interface for message in st.session_state.messages: with st.chat_message(message["role"]): st.markdown(message["content"], unsafe_allow_html=True) # User Input if prompt := st.chat_input("Ask about your documents..."): # Add user message st.session_state.messages.append({ "role": "user", "content": f"**{prompt}**" }) # Generate response with st.spinner("🔍 Analyzing documents..."): reply = generate_response( prompt, st.session_state.language, st.session_state.files ) # Add assistant response st.session_state.messages.append({ "role": "assistant", "content": reply }) # Rerun to update display st.rerun() # Apply CSS last inject_custom_css()