Spaces:
Sleeping
Sleeping
| import os | |
| from langchain_core.prompts import PromptTemplate | |
| from langchain_community.output_parsers.rail_parser import GuardrailsOutputParser | |
| from langchain_community.document_loaders import PyPDFLoader | |
| import google.generativeai as genai | |
| import gradio as gr | |
| # Function for initialization | |
| def initialize(pdf_file, question): | |
| try: | |
| # Access the uploaded file information from Gradio | |
| file_info = pdf_file | |
| # Check if a file was uploaded | |
| if file_info is not None: | |
| # Construct potential file path based on temporary directory and filename | |
| file_path = os.path.join("/tmp", file_info.name) # Adjust temporary directory if needed | |
| if os.path.exists(file_path): | |
| # Process the PDF | |
| pdf_loader = PyPDFLoader(file_path) | |
| pages = pdf_loader.load_and_split() | |
| processed_context = "\n".join(str(page.page_content) for page in pages[:30]) # Limit to first 30 pages | |
| # Configure Google Generative AI (replace with your API key) | |
| genai.configure(api_key=os.getenv("GOOGLE_API_KEY")) | |
| # Prompt template for formatting context and question | |
| prompt_template = """Answer the question as precise as possible using the provided context. If the answer is not contained in the context, say "answer not available in context" | |
| Context: | |
| {context} | |
| Question: | |
| {question} | |
| Answer: | |
| """ | |
| prompt = PromptTemplate(template=prompt_template, input_variables=["context", "question"]) | |
| # Load the GeminiPro model | |
| model = genai.GenerativeModel('gemini-pro') | |
| # Option 1: Using GeminiPro's Text Generation (if applicable) | |
| # Check if the model has a 'generate' method (or similar) - adjust based on actual method | |
| if hasattr(model, 'generate'): | |
| # Process context and question (already done) | |
| # Generate answer using GeminiPro's generate method | |
| generated_answer = model.generate(prompt=prompt) # Replace with the appropriate method | |
| # Extract the answer (parse the output from 'generate') | |
| # ... (implementation depends on the model's output format) | |
| return generated_answer | |
| # Option 2: Alternative LLM Integration (if GeminiPro methods not suitable) | |
| # Replace this section with code using an alternative library/framework | |
| # for question answering (e.g., transformers, haystack) | |
| # Ensure the code integrates with your chosen LLM and handles context processing, | |
| # question answering, and answer extraction. | |
| # Example placeholder (replace with your actual implementation): | |
| # return "Alternative LLM integration not yet implemented." | |
| else: | |
| return "Error: The uploaded file could not be found." | |
| else: | |
| return "Error: No PDF file was uploaded." | |
| except Exception as e: | |
| return f"An error occurred: {e}" # Generic error handling | |
| # Create a Gradio interface | |
| interface = gr.Interface( | |
| fn=initialize, | |
| inputs=[ | |
| gr.File(label="Upload PDF"), # No need for 'type' argument | |
| gr.Textbox(label="Question") | |
| ], | |
| outputs="text", | |
| title="GeminiPro Q&A Bot", | |
| description="Ask questions about the uploaded PDF document.", | |
| ) | |
| # Launch the interface | |
| interface.launch() | |