Update app.py

app.py

@@ -1,19 +1,51 @@
+# Imports
+import base64
+import glob
+import json
+import math
+import mistune
+import openai
+import os
+import pytz
+import re
 import requests
 import streamlit as st
-import os
+import textract
+import time
+import zipfile
+from audio_recorder_streamlit import audio_recorder
+from bs4 import BeautifulSoup
+from collections import deque
+from datetime import datetime
+from dotenv import load_dotenv
 from huggingface_hub import InferenceClient
+from io import BytesIO
+from langchain.chat_models import ChatOpenAI
+from langchain.chains import ConversationalRetrievalChain
+from langchain.embeddings import OpenAIEmbeddings
+from langchain.memory import ConversationBufferMemory
+from langchain.text_splitter import CharacterTextSplitter
+from langchain.vectorstores import FAISS
+from openai import ChatCompletion
+from PyPDF2 import PdfReader
+from templates import bot_template, css, user_template
+from xml.etree import ElementTree as ET
 
-API_URL = 'https://qe55p8afio98s0u3.us-east-1.aws.endpoints.huggingface.cloud'
+# Constants
+API_URL = 'https://qe55p8afio98s0u3.us-east-1.aws.endpoints.huggingface.cloud'  # Dr Llama
 API_KEY = os.getenv('API_KEY')
-
 headers = {
     "Authorization": f"Bearer {API_KEY}",
     "Content-Type": "application/json"
 }
-
-# Prompt Set of Examples:
 prompt = f"Write instructions to teach anyone to write a discharge plan. List the entities, features and relationships to CCDA and FHIR objects in boldface."
+# page config and sidebar declares up front allow all other functions to see global class variables
+st.set_page_config(page_title="GPT Streamlit Document Reasoner", layout="wide")
+
+# UI Controls
+should_save = st.sidebar.checkbox("💾 Save", value=True)
 
+# Functions
 def StreamLLMChatResponse(prompt):
     endpoint_url = API_URL
     hf_token = API_KEY
@@ -39,7 +71,6 @@ def StreamLLMChatResponse(prompt):
         collected_chunks.append(r.token.text)
         chunk_message = r.token.text
         collected_messages.append(chunk_message)
-
         try:
             report.append(r.token.text)
             if len(r.token.text) > 0:
@@ -56,52 +87,6 @@ def query(payload):
 def get_output(prompt):
     return query({"inputs": prompt})
 
-
-
-
-
-import streamlit as st
-import openai
-import os
-import base64
-import glob
-import json
-import mistune
-import pytz
-import math
-import requests
-import time
-import re
-import textract
-import zipfile  # New import for zipping files
-
-
-from datetime import datetime
-from openai import ChatCompletion
-from xml.etree import ElementTree as ET
-from bs4 import BeautifulSoup
-from collections import deque
-from audio_recorder_streamlit import audio_recorder
-from dotenv import load_dotenv
-from PyPDF2 import PdfReader
-from langchain.text_splitter import CharacterTextSplitter
-from langchain.embeddings import OpenAIEmbeddings
-from langchain.vectorstores import FAISS
-from langchain.chat_models import ChatOpenAI
-from langchain.memory import ConversationBufferMemory
-from langchain.chains import ConversationalRetrievalChain
-from templates import css, bot_template, user_template
-
-# page config and sidebar declares up front allow all other functions to see global class variables
-st.set_page_config(page_title="GPT Streamlit Document Reasoner", layout="wide")
-should_save = st.sidebar.checkbox("💾 Save", value=True)
-
-def generate_filename_old(prompt, file_type):
-    central = pytz.timezone('US/Central')
-    safe_date_time = datetime.now(central).strftime("%m%d_%H%M")  # Date and time DD-HHMM
-    safe_prompt = "".join(x for x in prompt if x.isalnum())[:90]  # Limit file name size and trim whitespace
-    return f"{safe_date_time}_{safe_prompt}.{file_type}"  # Return a safe file name
-
 def generate_filename(prompt, file_type):
     central = pytz.timezone('US/Central')
     safe_date_time = datetime.now(central).strftime("%m%d_%H%M")
@@ -121,10 +106,7 @@ def transcribe_audio(openai_key, file_path, model):
         st.write(response.json())
         chatResponse = chat_with_model(response.json().get('text'), '') # *************************************
         transcript = response.json().get('text')
-        #st.write('Responses:')
-        #st.write(chatResponse)
         filename = generate_filename(transcript, 'txt')
-        #create_file(filename, transcript, chatResponse)
         response = chatResponse
         user_prompt = transcript
         create_file(filename, user_prompt, response, should_save)
@@ -147,47 +129,21 @@ def save_and_play_audio(audio_recorder):
 def create_file(filename, prompt, response, should_save=True):
     if not should_save:
         return
-
-    # Step 2: Extract base filename without extension
     base_filename, ext = os.path.splitext(filename)
-
-    # Step 3: Check if the response contains Python code
     has_python_code = bool(re.search(r"```python([\s\S]*?)```", response))
-
-    # Step 4: Write files based on type
     if ext in ['.txt', '.htm', '.md']:
-        # Create Prompt file
         with open(f"{base_filename}-Prompt.txt", 'w') as file:
             file.write(prompt)
-        
-        # Create Response file
         with open(f"{base_filename}-Response.md", 'w') as file:
             file.write(response)
-
-        # Create Code file if Python code is present
         if has_python_code:
-            # Extract Python code from the response
             python_code = re.findall(r"```python([\s\S]*?)```", response)[0].strip()
-
             with open(f"{base_filename}-Code.py", 'w') as file:
                 file.write(python_code)
-
-
-def create_file_old(filename, prompt, response, should_save=True):
-    if not should_save:
-        return
-    if filename.endswith(".txt"):
-        with open(filename, 'w') as file:
-            file.write(f"{prompt}\n{response}")
-    elif filename.endswith(".htm"):
-        with open(filename, 'w') as file:
-            file.write(f"{prompt}   {response}")
-    elif filename.endswith(".md"):
-        with open(filename, 'w') as file:
-            file.write(f"{prompt}\n\n{response}")
             
 def truncate_document(document, length):
     return document[:length]
+
 def divide_document(document, max_length):
     return [document[i:i+max_length] for i in range(0, len(document), max_length)]
 
@@ -252,35 +208,23 @@ def chat_with_model(prompt, document_section, model_choice='gpt-3.5-turbo'):
     conversation.append({'role': 'user', 'content': prompt})
     if len(document_section)>0:
         conversation.append({'role': 'assistant', 'content': document_section})
-        
     start_time = time.time()
     report = []
     res_box = st.empty()
     collected_chunks = []
     collected_messages = []
-
-    for chunk in openai.ChatCompletion.create(
-        model='gpt-3.5-turbo',
-        messages=conversation,
-        temperature=0.5,
-        stream=True  
-    ):
-        
-        collected_chunks.append(chunk)  # save the event response
-        chunk_message = chunk['choices'][0]['delta']  # extract the message
-        collected_messages.append(chunk_message)  # save the message
-        
+    for chunk in openai.ChatCompletion.create(model='gpt-3.5-turbo', messages=conversation, temperature=0.5, stream=True):
+        collected_chunks.append(chunk)  
+        chunk_message = chunk['choices'][0]['delta']  
+        collected_messages.append(chunk_message) 
         content=chunk["choices"][0].get("delta",{}).get("content")
-        
         try:
             report.append(content)
             if len(content) > 0:
                 result = "".join(report).strip()
-                #result = result.replace("\n", "")        
                 res_box.markdown(f'*{result}*') 
         except:
             st.write(' ')
-        
     full_reply_content = ''.join([m.get('content', '') for m in collected_messages])
     st.write("Elapsed time:")
     st.write(time.time() - start_time)
@@ -295,7 +239,6 @@ def chat_with_file_contents(prompt, file_content, model_choice='gpt-3.5-turbo'):
     return response['choices'][0]['message']['content']
 
 def extract_mime_type(file):
-    # Check if the input is a string
     if isinstance(file, str):
         pattern = r"type='(.*?)'"
         match = re.search(pattern, file)
@@ -303,15 +246,11 @@ def extract_mime_type(file):
             return match.group(1)
         else:
             raise ValueError(f"Unable to extract MIME type from {file}")
-    # If it's not a string, assume it's a streamlit.UploadedFile object
     elif isinstance(file, streamlit.UploadedFile):
         return file.type
     else:
         raise TypeError("Input should be a string or a streamlit.UploadedFile object")
 
-from io import BytesIO
-import re
-
 def extract_file_extension(file):
     # get the file name directly from the UploadedFile object
     file_name = file.name
@@ -326,10 +265,7 @@ def pdf2txt(docs):
     text = ""
     for file in docs:
         file_extension = extract_file_extension(file)
-        # print the file extension
         st.write(f"File type extension: {file_extension}")
-
-        # read the file according to its extension
         try:
             if file_extension.lower() in ['py', 'txt', 'html', 'htm', 'xml', 'json']:
                 text += file.getvalue().decode('utf-8')
@@ -340,20 +276,6 @@ def pdf2txt(docs):
                     text += pdf.pages[page].extract_text() # new PyPDF2 syntax
         except Exception as e:
             st.write(f"Error processing file {file.name}: {e}")
-
-    return text
-
-def pdf2txt_old(pdf_docs):
-    st.write(pdf_docs)
-    for file in pdf_docs:
-        mime_type = extract_mime_type(file)
-        st.write(f"MIME type of file: {mime_type}")
-    
-    text = ""
-    for pdf in pdf_docs:
-        pdf_reader = PdfReader(pdf)
-        for page in pdf_reader.pages:
-            text += page.extract_text()
     return text
 
 def txt2chunks(text):
@@ -376,13 +298,10 @@ def process_user_input(user_question):
     for i, message in enumerate(st.session_state.chat_history):
         template = user_template if i % 2 == 0 else bot_template
         st.write(template.replace("{{MSG}}", message.content), unsafe_allow_html=True)
-        # Save file output from PDF query results
         filename = generate_filename(user_question, 'txt')
-        #create_file(filename, user_question, message.content)
         response = message.content
         user_prompt = user_question
         create_file(filename, user_prompt, response, should_save)       
-        #st.sidebar.markdown(get_table_download_link(filename), unsafe_allow_html=True)
 
 def divide_prompt(prompt, max_length):
     words = prompt.split()
@@ -391,78 +310,53 @@ def divide_prompt(prompt, max_length):
     current_length = 0
     for word in words:
         if len(word) + current_length <= max_length:
-            current_length += len(word) + 1  # Adding 1 to account for spaces
+            current_length += len(word) + 1 
             current_chunk.append(word)
         else:
             chunks.append(' '.join(current_chunk))
             current_chunk = [word]
             current_length = len(word)
-    chunks.append(' '.join(current_chunk))  # Append the final chunk
+    chunks.append(' '.join(current_chunk))
     return chunks
 
 def create_zip_of_files(files):
-    """
-    Create a zip file from a list of files.
-    """
     zip_name = "all_files.zip"
     with zipfile.ZipFile(zip_name, 'w') as zipf:
         for file in files:
             zipf.write(file)
     return zip_name
 
-
 def get_zip_download_link(zip_file):
-    """
-    Generate a link to download the zip file.
-    """
     with open(zip_file, 'rb') as f:
         data = f.read()
     b64 = base64.b64encode(data).decode()
     href = f'<a href="data:application/zip;base64,{b64}" download="{zip_file}">Download All</a>'
     return href
 
-
-
 def main():
     st.title("Medical Llama Test Bench with Inference Endpoints Llama 7B")
     prompt = f"Write instructions to teach anyone to write a discharge plan. List the entities, features and relationships to CCDA and FHIR objects in boldface."
     example_input = st.text_input("Enter your example text:", value=prompt)
-
     if st.button("Run Prompt With Dr Llama"):
         try:
             StreamLLMChatResponse(example_input)
         except:
             st.write('Dr. Llama is asleep.  Starting up now on A10 - please give 5 minutes then retry as KEDA scales up from zero to activate running container(s).')
-
-# clip ---
-
     openai.api_key = os.getenv('OPENAI_KEY')
-
-    # File type for output, model choice
     menu = ["txt", "htm", "xlsx", "csv", "md", "py"]
     choice = st.sidebar.selectbox("Output File Type:", menu)
     model_choice = st.sidebar.radio("Select Model:", ('gpt-3.5-turbo', 'gpt-3.5-turbo-0301'))
-
-    # Audio, transcribe, GPT:
     filename = save_and_play_audio(audio_recorder)
     if filename is not None:
         transcription = transcribe_audio(openai.api_key, filename, "whisper-1")
         st.sidebar.markdown(get_table_download_link(filename), unsafe_allow_html=True)
         filename = None
-
-    # prompt interfaces
     user_prompt = st.text_area("Enter prompts, instructions & questions:", '', height=100)
-
-    # file section interface for prompts against large documents as context
     collength, colupload = st.columns([2,3])  # adjust the ratio as needed
     with collength:
         max_length = st.slider("File section length for large files", min_value=1000, max_value=128000, value=12000, step=1000)
     with colupload:
         uploaded_file = st.file_uploader("Add a file for context:", type=["pdf", "xml", "json", "xlsx", "csv", "html", "htm", "md", "txt"])
-
-
-    # Document section chat
-        
     document_sections = deque()
     document_responses = {}
     if uploaded_file is not None:
@@ -480,54 +374,36 @@ def main():
             else:
                 if st.button(f"Chat about Section {i+1}"):
                     st.write('Reasoning with your inputs...')
-                    response = chat_with_model(user_prompt, section, model_choice) # *************************************
+                    response = chat_with_model(user_prompt, section, model_choice)
                     st.write('Response:')
                     st.write(response)
                     document_responses[i] = response
                     filename = generate_filename(f"{user_prompt}_section_{i+1}", choice)
                     create_file(filename, user_prompt, response, should_save)
                     st.sidebar.markdown(get_table_download_link(filename), unsafe_allow_html=True)
-
     if st.button('💬 Chat'):
         st.write('Reasoning with your inputs...')
-        
-        #response = chat_with_model(user_prompt, ''.join(list(document_sections,)), model_choice) # *************************************
-
-        # Divide the user_prompt into smaller sections
         user_prompt_sections = divide_prompt(user_prompt, max_length)
         full_response = ''
         for prompt_section in user_prompt_sections:
-            # Process each section with the model
             response = chat_with_model(prompt_section, ''.join(list(document_sections)), model_choice)
             full_response += response + '\n'  # Combine the responses
-        
-        #st.write('Response:')
-        #st.write(full_response)
-
         response = full_response
         st.write('Response:')
         st.write(response)
-        
         filename = generate_filename(user_prompt, choice)
         create_file(filename, user_prompt, response, should_save)
         st.sidebar.markdown(get_table_download_link(filename), unsafe_allow_html=True)
-
     all_files = glob.glob("*.*")
     all_files = [file for file in all_files if len(os.path.splitext(file)[0]) >= 20]  # exclude files with short names
     all_files.sort(key=lambda x: (os.path.splitext(x)[1], x), reverse=True)  # sort by file type and file name in descending order
-
-    # Added "Delete All" button
     if st.sidebar.button("🗑 Delete All"):
         for file in all_files:
             os.remove(file)
         st.experimental_rerun()
-
-    # Added "Download All" button
     if st.sidebar.button("⬇️ Download All"):
         zip_file = create_zip_of_files(all_files)
         st.sidebar.markdown(get_zip_download_link(zip_file), unsafe_allow_html=True)
-        
-    # Sidebar of Files Saving History and surfacing files as context of prompts and responses
     file_contents=''
     next_action=''
     for file in all_files:
@@ -553,7 +429,6 @@ def main():
             if st.button("🗑", key="delete_"+file):
                 os.remove(file)
                 st.experimental_rerun()
-                
     if len(file_contents) > 0:
         if next_action=='open':
             file_content_area = st.text_area("File Contents:", file_contents, height=500)
@@ -565,19 +440,14 @@ def main():
             response = chat_with_model(user_prompt, file_contents, model_choice)
             filename = generate_filename(file_contents, choice)
             create_file(filename, user_prompt, response, should_save)
-
             st.experimental_rerun()
-            #st.sidebar.markdown(get_table_download_link(filename), unsafe_allow_html=True)
-                
 
 load_dotenv()
 st.write(css, unsafe_allow_html=True)
-
 st.header("Chat with documents :books:")
 user_question = st.text_input("Ask a question about your documents:")
 if user_question:
     process_user_input(user_question)
-
 with st.sidebar:
     st.subheader("Your documents")
     docs = st.file_uploader("import documents", accept_multiple_files=True)
@@ -591,11 +461,6 @@ with st.sidebar:
             st.markdown('# AI Search Index of Length:' + length + ' Created.')  # add timing
             filename = generate_filename(raw, 'txt')
             create_file(filename, raw, '', should_save)
-            #create_file(filename, raw, '')
-
-
-
-
 
 if __name__ == "__main__":
     main()