Duplicate from lekkalar/chatgpt-for-pdfs-without-chat-history

Co-authored-by: radhika lekkala <lekkalar@users.noreply.huggingface.co>

.gitattributes

@@ -0,0 +1,34 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,13 @@
+---
+title: ChatGPT For PDFs
+emoji: 👁
+colorFrom: indigo
+colorTo: gray
+sdk: gradio
+sdk_version: 3.33.1
+app_file: app.py
+pinned: false
+duplicated_from: lekkalar/chatgpt-for-pdfs-without-chat-history
+---
+
+Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -0,0 +1,192 @@
+import gradio as gr
+import os
+import time
+import pandas as pd
+
+
+from langchain.document_loaders import OnlinePDFLoader #for laoding the pdf
+from langchain.embeddings import OpenAIEmbeddings # for creating embeddings
+from langchain.vectorstores import Chroma # for the vectorization part
+from langchain.chains import RetrievalQA # for conversing with chatGPT
+from langchain.chat_models import ChatOpenAI # the LLM model we'll use (ChatGPT)
+from langchain import PromptTemplate
+
+def load_pdf_and_generate_embeddings(pdf_doc, open_ai_key, relevant_pages):
+    if openai_key is not None:
+        os.environ['OPENAI_API_KEY'] = open_ai_key
+        #Load the pdf file
+        loader = OnlinePDFLoader(pdf_doc.name)
+        pages = loader.load_and_split()
+        
+        #Create an instance of OpenAIEmbeddings, which is responsible for generating embeddings for text
+        embeddings = OpenAIEmbeddings()
+
+        pages_to_be_loaded =[]
+
+        if relevant_pages:
+            page_numbers = relevant_pages.split(",")
+            if len(page_numbers) != 0:
+                for page_number in page_numbers:
+                    if page_number.isdigit():
+                        pageIndex = int(page_number)-1
+                        if pageIndex >=0 and pageIndex <len(pages):
+                            pages_to_be_loaded.append(pages[pageIndex])
+        #In the scenario where none of the page numbers supplied exist in the PDF, we will revert to using the entire PDF.
+        if len(pages_to_be_loaded) ==0:
+            pages_to_be_loaded = pages.copy()
+            
+             
+        #To create a vector store, we use the Chroma class, which takes the documents (pages in our case) and the embeddings instance
+        vectordb = Chroma.from_documents(pages_to_be_loaded, embedding=embeddings)
+        
+        #Finally, we create the bot using the RetrievalQA class
+        global pdf_qa
+
+        prompt_template = """Use the following pieces of context to answer the question at the end. If you do not know the answer, just return N/A. If you encounter a date, return it in mm/dd/yyyy format.
+
+        {context}
+
+        Question: {question}
+        Return just the answer :"""
+        PROMPT = PromptTemplate(template=prompt_template, input_variables=["context", "question"])
+        chain_type_kwargs = {"prompt": PROMPT}
+        pdf_qa = RetrievalQA.from_chain_type(llm=ChatOpenAI(temperature=0, model_name="gpt-4"),chain_type="stuff", retriever=vectordb.as_retriever(search_kwargs={"k": 5}), chain_type_kwargs=chain_type_kwargs, return_source_documents=False)
+                
+        return "Ready"
+    else:
+        return "Please provide an OpenAI gpt-4 API key"
+        
+
+def answer_predefined_questions(document_type):
+    
+    if document_type == "Deed of Trust":
+        #Create a list of questions around the relevant fields of a Deed of Trust(DOT) document
+        query1 = "what is the Loan Number?"
+        field1 = "Loan Number"
+        query2 = "Who is the Borrower?"
+        field2 = "Borrower"
+        query3 = "what is the Case Number?"
+        field3 = "Case Number"
+        query4 = "what is the Mortgage Identification number?"
+        field4 = "MIN Number"
+        query5 = "DOT signed date?"
+        field5 = "Signed Date"
+        query6 = "Who is the Lender?"
+        field6 = "Lender"
+        query7 = "what is the VA/FHA Number?"
+        field7 = "VA/FHA Number"
+        query8 = "Who is the Co-Borrower?"
+        field8 = "Co-Borrower"
+        query9 = "What is the property type - single family, multi family?"
+        field9 = "Property Type"
+        query10 = "what is the Property Address?"
+        field10 = "Property Address"
+        query11 = "In what County is the property located?"
+        field11 = "Property County"
+        query12 = "what is the Electronically recorded date"
+        field12 = "Electronic Recording Date"
+
+    
+        
+    elif document_type == "Transmittal Summary":
+        #Create a list of questions around the relevant fields of a TRANSMITTAL SUMMARY document
+        query1 = "Who is the Borrower?"
+        field1 = "Borrower"
+        query2 = "what is the Property Address?"
+        field2 = "Property Address"
+        query3 = "what is the Loan Term?"
+        field3 = "Loan Term"
+        query4 = "What is the Base Income?"
+        field4 = "Base Income"
+        query5 = "what is the Borrower's SSN?"
+        field5 = "Borrower's SSN"
+        query6 = "Who is the Co-Borrower?"
+        field6 = "Co-Borrower"
+        query7 = "What is the Original Loan Amount?"
+        field7 = "Original Loan Amount"
+        query8 = "What is the Initial P&I payment?"
+        field8 = "Initial P&I payment"
+        query9 = "What is the Co-Borrower's SSN?"
+        field9 = "Co-Borrower’s SSN"
+        query10 = "Number of units?"
+        field10 = "Units#"
+        query11 = "Who is the Seller?"
+        field11 = "Seller"
+        query12 = "Document signed date?"
+        field12 = "Signed Date"
+
+
+
+    else:
+        return "Please choose your Document Type"
+
+    queryList = [query1, query2, query3, query4, query5, query6, query7, query8, query9, query10, query11,query12]
+    fieldList = [field1, field2, field3, field4, field5, field6, field7, field8, field9, field10, field11,field12]
+    responseList =[]
+
+    i = 0
+    while i < len(queryList):
+        question = queryList[i]
+        responseList.append(pdf_qa.run(question))
+        i = i+1
+
+    return pd.DataFrame({"Field": [fieldList[0],fieldList[1],fieldList[2],fieldList[3],fieldList[4],fieldList[5],fieldList[6],fieldList[7],fieldList[8],fieldList[9],fieldList[10],fieldList[11]], 
+                         "Question to gpt-4": [queryList[0],queryList[1],queryList[2],queryList[3],queryList[4],queryList[5],queryList[6],queryList[7],queryList[8],queryList[9],queryList[10],queryList[11]], 
+                         "Response from gpt-4": [responseList[0],responseList[1],responseList[2],responseList[3],responseList[4],responseList[5],responseList[6],responseList[7],responseList[8],responseList[9],responseList[10],responseList[11]]})
+        
+
+    
+def answer_query(query):
+    question = query
+    return pdf_qa.run(question)
+    
+
+css="""
+#col-container {max-width: 700px; margin-left: auto; margin-right: auto;}
+"""
+
+title = """
+<div style="text-align: center;max-width: 700px;">
+    <h1>Chatbot for PDFs - GPT-4</h1>
+    <p style="text-align: center;">Upload a .PDF, click the "Upload PDF and generate embeddings" button, <br />
+    Wait for the Status to show Ready. You can chose to get answers to the pre-defined question set OR ask your own question <br />
+    The app is built on GPT-4 and leverages PromptTemplate</p>
+</div>
+"""
+
+with gr.Blocks(css=css,theme=gr.themes.Monochrome()) as demo:
+    with gr.Column(elem_id="col-container"):
+        gr.HTML(title)
+    
+    with gr.Column():
+        openai_key = gr.Textbox(label="Your GPT-4 OpenAI API key", type="password")
+        pdf_doc = gr.File(label="Load a pdf",file_types=['.pdf'],type='file')
+        relevant_pages = gr.Textbox(label="*Optional - List comma separated page numbers to load or leave this field blank to use the entire PDF")
+        
+        with gr.Row():
+            status = gr.Textbox(label="Status", placeholder="", interactive=False)
+            load_pdf = gr.Button("Upload PDF and generate embeddings").style(full_width=False)
+            
+        with gr.Row():
+            document_type = gr.Radio(['Deed of Trust', 'Transmittal Summary'], label="Select the Document Type")
+            answers = gr.Dataframe(label="Answers to Predefined Question set")
+            answers_for_predefined_question_set = gr.Button("Get gpt-4 answers to pre-defined question set").style(full_width=False)
+              
+        with gr.Row():
+            input = gr.Textbox(label="Type in your question")
+            output = gr.Textbox(label="Answer")
+            submit_query = gr.Button("Submit your own question to gpt-4").style(full_width=False)
+        
+ 
+    load_pdf.click(load_pdf_and_generate_embeddings, inputs=[pdf_doc, openai_key, relevant_pages], outputs=status)
+
+    answers_for_predefined_question_set.click(answer_predefined_questions, document_type, answers)
+        
+    submit_query.click(answer_query,input,output)
+
+
+demo.launch()
+
+
+
+

requirements.txt

@@ -0,0 +1,6 @@
+openai
+tiktoken
+chromadb
+langchain
+unstructured
+unstructured[local-inference]