Create app.py

app.py

@@ -0,0 +1,130 @@
+import gradio as gr
+from langchain_community.document_loaders import PyPDFLoader
+from langchain_community.vectorstores import FAISS
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain.chains import RetrievalQA
+from langchain_community.llms import HuggingFaceHub
+from langchain_community.embeddings import HuggingFaceInferenceAPIEmbeddings
+
+# You can use this section to suppress warnings generated by your code:
+def warn(*args, **kwargs):
+    pass
+import warnings
+warnings.warn = warn
+warnings.filterwarnings('ignore')
+
+# Set your Hugging Face API token here.
+# For deployment on Hugging Face, you can set this as an environment variable.
+import os
+os.environ["HUGGINGFACEHUB_API_TOKEN"] = "hf_YOUR_HUGGINGFACE_TOKEN"
+
+## LLM - Using an open-source model from Hugging Face
+def get_llm():
+    """
+    Initializes and returns a Hugging Face Hub LLM model.
+    Using a conversational model suitable for legal advice.
+    """
+    repo_id = "mistralai/Mixtral-8x7B-Instruct-v0.1"
+    llm = HuggingFaceHub(
+        repo_id=repo_id,
+        model_kwargs={"temperature": 0.1, "max_length": 512}
+    )
+    return llm
+
+## Document loader
+def document_loader(file_path):
+    """
+    Loads a PDF document from the given file path.
+    """
+    loader = PyPDFLoader(file_path)
+    loaded_document = loader.load()
+    return loaded_document
+
+## Text splitter
+def text_splitter(data):
+    """
+    Splits the loaded document into smaller chunks for processing.
+    """
+    text_splitter = RecursiveCharacterTextSplitter(
+        chunk_size=1000,
+        chunk_overlap=200,
+        length_function=len,
+    )
+    chunks = text_splitter.split_documents(data)
+    return chunks
+
+## Vector db and Embedding model
+def vector_database(chunks):
+    """
+    Creates a FAISS vector database from the document chunks using a
+    Hugging Face embeddings model.
+    """
+    # Using a sentence-transformer model from Hugging Face for embeddings
+    embedding_model = HuggingFaceInferenceAPIEmbeddings(
+        api_key=os.environ["HUGGINGFACEHUB_API_TOKEN"],
+        model_name="sentence-transformers/all-MiniLM-L6-v2"
+    )
+    vectordb = FAISS.from_documents(chunks, embedding_model)
+    return vectordb
+
+## Retriever
+def retriever(file_path):
+    """
+    Loads, splits, and creates a retriever for the document.
+    """
+    splits = document_loader(file_path)
+    chunks = text_splitter(splits)
+    vectordb = vector_database(chunks)
+    retriever = vectordb.as_retriever()
+    return retriever
+
+## QA Chain
+def retriever_qa(file, query):
+    """
+    Sets up a RetrievalQA chain to answer questions based on the document.
+    """
+    llm = get_llm()
+    retriever_obj = retriever(file)
+    
+    # Custom prompt to act as a conversational legal advisor
+    prompt_template = f"""
+    You are a friendly and professional legal advisor. Your goal is to provide concise and contextual legal advice based on the provided document.
+    Do not give verbatim answers. Instead, analyze the relevant text and respond in a conversational manner.
+    
+    Context:
+    {file}
+    
+    Question: {query}
+    
+    Legal Advisor's Answer:
+    """
+    
+    qa = RetrievalQA.from_chain_type(
+        llm=llm,
+        chain_type="stuff",
+        retriever=retriever_obj,
+        return_source_documents=True,
+    )
+    
+    # Using a custom prompt template for the LLM
+    response = qa.invoke({"query": prompt_template})
+    
+    # Extract the contextual response from the full LLM output
+    result_text = response['result']
+    return result_text
+
+# Create Gradio interface
+rag_application = gr.Interface(
+    fn=retriever_qa,
+    allow_flagging="never",
+    inputs=[
+        gr.File(label="Upload PDF File", file_count="single", file_types=['.pdf'], type="filepath"),
+        gr.Textbox(label="Input Query", lines=2, placeholder="Type your question here...")
+    ],
+    outputs=gr.Textbox(label="Legal Advisor's Response"),
+    title="Nigerian Constitution Legal Advisor Chatbot",
+    description="Upload the Nigerian Constitution and ask me questions about it. I will provide a conversational and contextual response."
+)
+
+# Launch the app
+rag_application.launch(share=True)