awacke1's picture
Update app.py
78eba48 verified
import streamlit as st
import os
import json
from PIL import Image
from urllib.parse import quote # Ensure this import is included
import base64
import glob
import json
import math
import openai
import os
import pytz
import re
import requests
import textract
import time
import zipfile
import huggingface_hub
import dotenv
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 openai import ChatCompletion
from PyPDF2 import PdfReader
#from templates import bot_template, css, user_template -- pattern with content
from xml.etree import ElementTree as ET
import streamlit.components.v1 as components # Import Streamlit Components for HTML5
# Set page configuration with a title and favicon
st.set_page_config(
page_title="🌌🚀 Mixable AI - Voice Search",
page_icon="🌠",
layout="wide",
initial_sidebar_state="expanded",
menu_items={
'Get Help': 'https://huggingface.co/awacke1',
'Report a bug': "https://huggingface.co/spaces/awacke1/WebDataDownload",
'About': "# Midjourney: https://discord.com/channels/@me/997514686608191558"
}
)
if st.checkbox('Show Anatomy Table'):
st.markdown("""
## Anatomy Head to Toe Table with Body Organs Costly Conditions, Spending, CPT Codes and Frequency
| Table Num | Body Part | Organ/Part | Description | 📈 Costly Condition | 💰 Spending (billions) | CPT Range Start | CPT Range Finish | Frequency |
|-----------|------------------|----------------------|-------------------------------|------------------------------|------------------------|-----------------|------------------|----------------|
| 1 | 🧠 Head | 🧠 Brain | Controls mental processes | 😨 Anxiety & Depression | 210 | 90791 | 90899 | 1 in 5 |
| 2 | 👀 Eyes | 👁️ Optic Nerve | Vision | 👓 Cataracts | 10.7 | 92002 | 92499 | 1 in 6 (over 40 years) |
| 3 | 👂 Ears | 🐚 Cochlea | Hearing | 📢 Hearing Loss | 7.1 | 92502 | 92700 | 1 in 8 (over 12 years) |
| 4 | 👃 Nose | 👃 Olfactory Bulb | Smell | 🤧 Allergies | 25 | 31231 | 31294 | 1 in 3 |
| 5 | 👄 Mouth | 👅 Tongue | Taste | 🦷 Dental Issues | 130 | 00100 | 00192 | 1 in 2 |
| 6 | 🫁 Neck | 🦋 Thyroid | Metabolism | 🦠 Hypothyroidism | 3.1 | 60210 | 60271 | 1 in 20 |
| 7 | 💪 Upper Body | ❤️ Heart | Circulation | 💔 Heart Disease | 230 | 92920 | 93799 | 1 in 4 (over 65 years) |
| 8 | 💪 Upper Body | 🫁 Lungs | Respiration | 😷 Chronic Obstructive Pulmonary Disease | 70 | 94002 | 94799 | 1 in 20 (over 45 years) |
| 9 | 💪 Upper Body | 🍷 Liver | Detoxification | 🍺 Liver Disease | 40 | 47000 | 47999 | 1 in 10 |
| 10 | 💪 Upper Body | 🍹 Kidneys | Filtration | 🌊 Chronic Kidney Disease | 110 | 50010 | 50999 | 1 in 7 |
| 11 | 💪 Upper Body | 💉 Pancreas | Insulin secretion | 🍬 Diabetes | 327 | 48100 | 48999 | 1 in 10 |
| 12 | 💪 Upper Body | 🍽️ Stomach | Digestion | 🔥 Gastroesophageal Reflux Disease | 17 | 43200 | 43289 | 1 in 5 |
| 13 | 💪 Upper Body | 🛡️ Spleen | Immune functions | 🩸 Anemia | 5.6 | 38100 | 38199 | 1 in 6 |
| 14 | 💪 Upper Body | 🫀 Blood Vessels | Circulation of blood | 🚑 Hypertension | 55 | 40110 | 40599 | 1 in 3 |
| 15 | 🦵 Lower Body | 🍝 Colon | Absorption of water, minerals | 🌟 Colorectal Cancer | 14 | 45378 | 45378 | 1 in 23 |
| 16 | 🦵 Lower Body | 🚽 Bladder | Urine excretion | 💧 Urinary Incontinence | 8 | 51700 | 51798 | 1 in 4 (over 65 years) |
| 17 | 🦵 Lower Body | 💞 Reproductive Organs | Sex hormone secretion | 🎗️ Endometriosis | 22 | 56405 | 58999 | 1 in 10 (women) |
| 18 | 🦶 Feet | 🎯 Nerve endings | Balance and movement | 🤕 Peripheral Neuropathy | 19 | 95900 | 96004 | 1 in 30 |
| 19 | 🦶 Feet | 🌡️ Skin | Temperature regulation | 🌞 Skin Cancer | 8.1 | 96910 | 96999 | 1 in 5 |
| 20 | 🦶 Feet | 💪 Muscles | Movement and strength | 🏋️‍♂️ Musculoskeletal Disorders | 176 | 97110 | 97799 | 1 in 2 |
""")
body_map_data = {
"🧠 Central Nervous System": {
"Brain": ["Cognitive functions", "Emotion regulation", "Neural coordination"],
"Spinal Cord": ["Nerve signal transmission", "Reflex actions", "Connects brain to body"],
},
"👀 Sensory Organs": {
"Eyes": ["Vision", "Light perception", "Color differentiation"],
"Ears": ["Hearing", "Balance maintenance", "Sound localization"],
"Nose": ["Smell detection", "Olfactory signaling", "Air filtration"],
"Tongue": ["Taste perception", "Texture sensing", "Temperature feeling"],
"Skin": ["Touch sensation", "Temperature regulation", "Protection against pathogens"],
},
"🫁 Respiratory System": {
"Lungs": ["Gas exchange", "Oxygen intake", "Carbon dioxide expulsion"],
"Trachea": ["Airway protection", "Mucus secretion", "Cough reflex"],
},
"❤️ Circulatory System": {
"Heart": ["Blood pumping", "Circulatory regulation", "Oxygen and nutrients distribution"],
"Blood Vessels": ["Blood transport", "Nutrient delivery", "Waste removal"],
},
"🍽️ Digestive System": {
"Stomach": ["Food breakdown", "Enzyme secretion", "Nutrient digestion"],
"Intestines": ["Nutrient absorption", "Waste processing", "Microbiome hosting"],
},
"💪 Musculoskeletal System": {
"Bones": ["Structural support", "Protection of organs", "Mineral storage"],
"Muscles": ["Movement facilitation", "Posture maintenance", "Heat production"],
},
"🚽 Excretory System": {
"Kidneys": ["Waste filtration", "Water balance", "Electrolyte regulation"],
"Bladder": ["Urine storage", "Excretion control", "Toxin removal"],
},
"💞 Endocrine System": {
"Thyroid": ["Metabolic regulation", "Hormone secretion", "Energy management"],
"Adrenal Glands": ["Stress response", "Metabolism control", "Immune system regulation"],
},
"🧬 Reproductive System": {
"Male Reproductive Organs": ["Sperm production", "Sexual function", "Hormone synthesis"],
"Female Reproductive Organs": ["Egg production", "Fetus gestation", "Hormone regulation"],
},
"🩸 Immune System": {
"White Blood Cells": ["Pathogen defense", "Infection response", "Immunity maintenance"],
"Lymphatic System": ["Fluid balance", "Waste removal", "Antibody production"],
},
"🧘 Integrative Body Functions": {
"Sleep Regulation": ["Rest and recovery", "Memory consolidation", "Energy conservation"],
"Stress Management": ["Coping mechanisms", "Hormonal balance", "Emotional regulation"],
},
"🔬 Research and Innovations": {
"Genetic Studies": ["Disease predisposition", "Trait inheritance", "Gene therapy"],
"Biomedical Engineering": ["Medical devices", "Prosthetics design", "Healthcare technologies"],
},
"🎓 Education and Awareness": {
"Anatomy and Physiology": ["Body structure", "Function understanding", "Health education"],
"Public Health Initiatives": ["Disease prevention", "Health promotion", "Community wellness"],
},
}
# Ensure the directory for storing scores exists
score_dir = "scores"
os.makedirs(score_dir, exist_ok=True)
# Function to generate a unique key for each button, including an emoji
def generate_key(label, header, idx):
return f"{header}_{label}_{idx}_key"
# Function to increment and save score
def update_score(key, increment=1):
score_file = os.path.join(score_dir, f"{key}.json")
if os.path.exists(score_file):
with open(score_file, "r") as file:
score_data = json.load(file)
else:
score_data = {"clicks": 0, "score": 0}
score_data["clicks"] += 1
score_data["score"] += increment
with open(score_file, "w") as file:
json.dump(score_data, file)
return score_data["score"]
# Function to load score
def load_score(key):
score_file = os.path.join(score_dir, f"{key}.json")
if os.path.exists(score_file):
with open(score_file, "r") as file:
score_data = json.load(file)
return score_data["score"]
return 0
def search_glossary(query):
st.write('## ' + query)
all=""
st.write('## 🔍 Running with GPT.') # -------------------------------------------------------------------------------------------------
response = chat_with_model(query)
#st.write(response)
filename = generate_filename(query + ' --- ' + response, "md")
create_file(filename, query, response, should_save)
#st.write('## 🔍 Running with Llama.') # -------------------------------------------------------------------------------------------------
#response2 = StreamLLMChatResponse(query)
#st.write(response2)
filename_txt = generate_filename(query + ' --- ' + response2, "md")
create_file(filename_txt, query, response2, should_save)
all = '# Query: ' + query + '# Response: ' + response + '# Response2: ' + response2
filename_txt2 = generate_filename(query + ' --- ' + all, "md")
create_file(filename_txt2, query, all, should_save)
SpeechSynthesis(all)
return all
# Function to display the glossary in a structured format
def display_glossary(glossary, area):
if area in glossary:
st.subheader(f"📘 Glossary for {area}")
for game, terms in glossary[area].items():
st.markdown(f"### {game}")
for idx, term in enumerate(terms, start=1):
st.write(f"{idx}. {term}")
# Function to display the entire glossary in a grid format with links
def display_glossary_grid(body_map_data):
search_urls = {
"📖": lambda k: f"https://en.wikipedia.org/wiki/{quote(k)}",
"🔍": lambda k: f"https://www.google.com/search?q={quote(k)}",
"▶️": lambda k: f"https://www.youtube.com/results?search_query={quote(k)}",
"🔎": lambda k: f"https://www.bing.com/search?q={quote(k)}",
"🎲": lambda k: f"https://huggingface.co/spaces/awacke1/AI-ChatGPT-CPT-Body-Map-Cost?q={quote(k)}", # this url plus query!
}
for category, details in body_map_data.items():
st.write(f"### {category}")
cols = st.columns(len(details)) # Create dynamic columns based on the number of games
for idx, (game, terms) in enumerate(details.items()):
with cols[idx]:
st.markdown(f"#### {game}")
for term in terms:
links_md = ' '.join([f"[{emoji}]({url(term)})" for emoji, url in search_urls.items()])
st.markdown(f"{term} {links_md}", unsafe_allow_html=True)
game_emojis = {
"Dungeons and Dragons": "🐉",
"Call of Cthulhu": "🐙",
"GURPS": "🎲",
"Pathfinder": "🗺️",
"Kindred of the East": "🌅",
"Changeling": "🍃",
}
topic_emojis = {
"Core Rulebooks": "📚",
"Maps & Settings": "🗺️",
"Game Mechanics & Tools": "⚙️",
"Monsters & Adversaries": "👹",
"Campaigns & Adventures": "📜",
"Creatives & Assets": "🎨",
"Game Master Resources": "🛠️",
"Lore & Background": "📖",
"Character Development": "🧍",
"Homebrew Content": "🔧",
"General Topics": "🌍",
}
# Adjusted display_buttons_with_scores function
def display_buttons_with_scores():
for category, games in body_map_data.items():
category_emoji = topic_emojis.get(category, "🔍") # Default to search icon if no match
st.markdown(f"## {category_emoji} {category}")
for game, terms in games.items():
game_emoji = game_emojis.get(game, "🎮") # Default to generic game controller if no match
for term in terms:
key = f"{category}_{game}_{term}".replace(' ', '_').lower()
score = load_score(key)
if st.button(f"{game_emoji} {term} {score}", key=key):
update_score(key)
# Create a dynamic query incorporating emojis and formatting for clarity
query_prefix = f"{category_emoji} {game_emoji} **{game} - {category}:**"
# -----------------------------------------------------------------
# query_body = f"Create a detailed outline for **{term}** with subpoints highlighting key aspects, using emojis for visual engagement. Include step-by-step rules and boldface important entities and ruleset elements."
query_body = f"Create a streamlit python app.py that produces a detailed markdown outline and CSV dataset user interface with an outline for **{term}** with subpoints highlighting key aspects, using emojis for visual engagement. Include step-by-step rules and boldface important entities and ruleset elements."
response = search_glossary(query_prefix + query_body, body_map_data)
def fetch_wikipedia_summary(keyword):
# Placeholder function for fetching Wikipedia summaries
# In a real app, you might use requests to fetch from the Wikipedia API
return f"Summary for {keyword}. For more information, visit Wikipedia."
def create_search_url_youtube(keyword):
base_url = "https://www.youtube.com/results?search_query="
return base_url + keyword.replace(' ', '+')
def create_search_url_bing(keyword):
base_url = "https://www.bing.com/search?q="
return base_url + keyword.replace(' ', '+')
def create_search_url_wikipedia(keyword):
base_url = "https://www.wikipedia.org/search-redirect.php?family=wikipedia&language=en&search="
return base_url + keyword.replace(' ', '+')
def create_search_url_google(keyword):
base_url = "https://www.google.com/search?q="
return base_url + keyword.replace(' ', '+')
def display_images_and_wikipedia_summaries():
st.title('Gallery with Related Stories')
image_files = [f for f in os.listdir('.') if f.endswith('.png')]
if not image_files:
st.write("No PNG images found in the current directory.")
return
for image_file in image_files:
image = Image.open(image_file)
st.image(image, caption=image_file, use_column_width=True)
keyword = image_file.split('.')[0] # Assumes keyword is the file name without extension
# Display Wikipedia and Google search links
wikipedia_url = create_search_url_wikipedia(keyword)
google_url = create_search_url_google(keyword)
youtube_url = create_search_url_youtube(keyword)
bing_url = create_search_url_bing(keyword)
links_md = f"""
[Wikipedia]({wikipedia_url}) |
[Google]({google_url}) |
[YouTube]({youtube_url}) |
[Bing]({bing_url})
"""
st.markdown(links_md)
def get_all_query_params(key):
return st.query_params().get(key, [])
def clear_query_params():
st.query_params()
# Function to display content or image based on a query
def display_content_or_image(query):
# Check if the query matches any glossary term
for category, terms in transhuman_glossary.items():
for term in terms:
if query.lower() in term.lower():
st.subheader(f"Found in {category}:")
st.write(term)
return True # Return after finding and displaying the first match
# Check for an image match in a predefined directory (adjust path as needed)
image_dir = "images" # Example directory where images are stored
image_path = f"{image_dir}/{query}.png" # Construct image path with query
if os.path.exists(image_path):
st.image(image_path, caption=f"Image for {query}")
return True
# If no content or image is found
st.warning("No matching content or image found.")
return False
# 1. Constants and Top Level UI Variables
# My Inference API Copy
API_URL = 'https://qe55p8afio98s0u3.us-east-1.aws.endpoints.huggingface.cloud' # Dr Llama
# Meta's Original - Chat HF Free Version:
#API_URL = "https://api-inference.huggingface.co/models/meta-llama/Llama-2-7b-chat-hf"
API_KEY = os.getenv('API_KEY')
MODEL1="meta-llama/Llama-2-7b-chat-hf"
MODEL1URL="https://huggingface.co/meta-llama/Llama-2-7b-chat-hf"
HF_KEY = os.getenv('HF_KEY')
headers = {
"Authorization": f"Bearer {HF_KEY}",
"Content-Type": "application/json"
}
key = os.getenv('OPENAI_API_KEY')
prompt = f"Write instructions to teach discharge planning along with guidelines and patient education. List entities, features and relationships to CCDA and FHIR objects in boldface."
should_save = st.sidebar.checkbox("💾 Save", value=True, help="Save your session data.")
def SpeechSynthesis(result):
documentHTML5='''
<!DOCTYPE html>
<html>
<head>
<title>Read It Aloud</title>
<script type="text/javascript">
function readAloud() {
const text = document.getElementById("textArea").value;
const speech = new SpeechSynthesisUtterance(text);
window.speechSynthesis.speak(speech);
}
</script>
</head>
<body>
<h1>🔊 Read It Aloud</h1>
<textarea id="textArea" rows="10" cols="80">
'''
documentHTML5 = documentHTML5 + result
documentHTML5 = documentHTML5 + '''
</textarea>
<br>
<button onclick="readAloud()">🔊 Read Aloud</button>
</body>
</html>
'''
components.html(documentHTML5, width=1280, height=300)
#return result
# 3. Stream Llama Response
# @st.cache_resource
def StreamLLMChatResponse(prompt):
try:
endpoint_url = API_URL
hf_token = API_KEY
st.write('Running client ' + endpoint_url)
client = InferenceClient(endpoint_url, token=hf_token)
gen_kwargs = dict(
max_new_tokens=512,
top_k=30,
top_p=0.9,
temperature=0.2,
repetition_penalty=1.02,
stop_sequences=["\nUser:", "<|endoftext|>", "</s>"],
)
stream = client.text_generation(prompt, stream=True, details=True, **gen_kwargs)
report=[]
res_box = st.empty()
collected_chunks=[]
collected_messages=[]
allresults=''
for r in stream:
if r.token.special:
continue
if r.token.text in gen_kwargs["stop_sequences"]:
break
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:
result="".join(report).strip()
res_box.markdown(f'*{result}*')
except:
st.write('Stream llm issue')
SpeechSynthesis(result)
return result
except:
st.write('Llama model is asleep. Starting up now on A10 - please give 5 minutes then retry as KEDA scales up from zero to activate running container(s).')
# 4. Run query with payload
def query(payload):
response = requests.post(API_URL, headers=headers, json=payload)
st.markdown(response.json())
return response.json()
def get_output(prompt):
return query({"inputs": prompt})
# 5. Auto name generated output files from time and content
def generate_filename(prompt, file_type):
central = pytz.timezone('US/Central')
safe_date_time = datetime.now(central).strftime("%m%d_%H%M")
replaced_prompt = prompt.replace(" ", "_").replace("\n", "_")
safe_prompt = "".join(x for x in replaced_prompt if x.isalnum() or x == "_")[:255] # 255 is linux max, 260 is windows max
#safe_prompt = "".join(x for x in replaced_prompt if x.isalnum() or x == "_")[:45]
return f"{safe_date_time}_{safe_prompt}.{file_type}"
# 6. Speech transcription via OpenAI service
def transcribe_audio(openai_key, file_path, model):
openai.api_key = openai_key
OPENAI_API_URL = "https://api.openai.com/v1/audio/transcriptions"
headers = {
"Authorization": f"Bearer {openai_key}",
}
with open(file_path, 'rb') as f:
data = {'file': f}
st.write('STT transcript ' + OPENAI_API_URL)
response = requests.post(OPENAI_API_URL, headers=headers, files=data, data={'model': model})
if response.status_code == 200:
st.write(response.json())
chatResponse = chat_with_model(response.json().get('text'), '') # *************************************
transcript = response.json().get('text')
filename = generate_filename(transcript, 'txt')
response = chatResponse
user_prompt = transcript
create_file(filename, user_prompt, response, should_save)
return transcript
else:
st.write(response.json())
st.error("Error in API call.")
return None
# 7. Auto stop on silence audio control for recording WAV files
def save_and_play_audio(audio_recorder):
audio_bytes = audio_recorder(key='audio_recorder')
if audio_bytes:
filename = generate_filename("Recording", "wav")
with open(filename, 'wb') as f:
f.write(audio_bytes)
st.audio(audio_bytes, format="audio/wav")
return filename
return None
# 8. File creator that interprets type and creates output file for text, markdown and code
def create_file(filename, prompt, response, should_save=True):
if not should_save:
return
base_filename, ext = os.path.splitext(filename)
if ext in ['.txt', '.htm', '.md']:
with open(f"{base_filename}.md", 'w') as file:
try:
content = prompt.strip() + '\r\n' + response
file.write(content)
except:
st.write('.')
#has_python_code = re.search(r"```python([\s\S]*?)```", prompt.strip() + '\r\n' + response)
#has_python_code = bool(re.search(r"```python([\s\S]*?)```", prompt.strip() + '\r\n' + response))
#if has_python_code:
# 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)
# with open(f"{base_filename}.md", 'w') as file:
# content = prompt.strip() + '\r\n' + response
# file.write(content)
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)]
# 9. Sidebar with UI controls to review and re-run prompts and continue responses
@st.cache_resource
def get_table_download_link(file_path):
with open(file_path, 'r') as file:
data = file.read()
b64 = base64.b64encode(data.encode()).decode()
file_name = os.path.basename(file_path)
ext = os.path.splitext(file_name)[1] # get the file extension
if ext == '.txt':
mime_type = 'text/plain'
elif ext == '.py':
mime_type = 'text/plain'
elif ext == '.xlsx':
mime_type = 'text/plain'
elif ext == '.csv':
mime_type = 'text/plain'
elif ext == '.htm':
mime_type = 'text/html'
elif ext == '.md':
mime_type = 'text/markdown'
elif ext == '.wav':
mime_type = 'audio/wav'
else:
mime_type = 'application/octet-stream' # general binary data type
href = f'<a href="data:{mime_type};base64,{b64}" target="_blank" download="{file_name}">{file_name}</a>'
return href
def CompressXML(xml_text):
root = ET.fromstring(xml_text)
for elem in list(root.iter()):
if isinstance(elem.tag, str) and 'Comment' in elem.tag:
elem.parent.remove(elem)
return ET.tostring(root, encoding='unicode', method="xml")
# 10. Read in and provide UI for past files
@st.cache_resource
def read_file_content(file,max_length):
if file.type == "application/json":
content = json.load(file)
return str(content)
elif file.type == "text/html" or file.type == "text/htm":
content = BeautifulSoup(file, "html.parser")
return content.text
elif file.type == "application/xml" or file.type == "text/xml":
tree = ET.parse(file)
root = tree.getroot()
xml = CompressXML(ET.tostring(root, encoding='unicode'))
return xml
elif file.type == "text/markdown" or file.type == "text/md":
md = mistune.create_markdown()
content = md(file.read().decode())
return content
elif file.type == "text/plain":
return file.getvalue().decode()
else:
return ""
# 11. Chat with GPT - Caution on quota
@st.cache_resource
def chat_with_model(prompt, document_section='', model_choice='gpt-3.5-turbo'):
model = model_choice
conversation = [{'role': 'system', 'content': 'You are a helpful assistant.'}]
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 = []
st.write('LLM stream ' + 'gpt-3.5-turbo')
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()
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)
return full_reply_content
def extract_mime_type(file):
if isinstance(file, str):
pattern = r"type='(.*?)'"
match = re.search(pattern, file)
if match:
return match.group(1)
else:
raise ValueError(f"Unable to extract MIME type from {file}")
elif isinstance(file, streamlit.UploadedFile):
return file.type
else:
raise TypeError("Input should be a string or a streamlit.UploadedFile object")
def extract_file_extension(file):
# get the file name directly from the UploadedFile object
file_name = file.name
pattern = r".*?\.(.*?)$"
match = re.search(pattern, file_name)
if match:
return match.group(1)
else:
raise ValueError(f"Unable to extract file extension from {file_name}")
# Normalize input as text from PDF and other formats
@st.cache_resource
def pdf2txt(docs):
text = ""
for file in docs:
file_extension = extract_file_extension(file)
st.write(f"File type extension: {file_extension}")
if file_extension.lower() in ['py', 'txt', 'html', 'htm', 'xml', 'json']:
text += file.getvalue().decode('utf-8')
elif file_extension.lower() == 'pdf':
from PyPDF2 import PdfReader
pdf = PdfReader(BytesIO(file.getvalue()))
for page in range(len(pdf.pages)):
text += pdf.pages[page].extract_text() # new PyPDF2 syntax
return text
def txt2chunks(text):
text_splitter = CharacterTextSplitter(separator="\n", chunk_size=1000, chunk_overlap=200, length_function=len)
return text_splitter.split_text(text)
# Vector Store using FAISS
@st.cache_resource
def vector_store(text_chunks):
embeddings = OpenAIEmbeddings(openai_api_key=key)
return FAISS.from_texts(texts=text_chunks, embedding=embeddings)
# Memory and Retrieval chains
@st.cache_resource
def get_chain(vectorstore):
llm = ChatOpenAI()
memory = ConversationBufferMemory(memory_key='chat_history', return_messages=True)
return ConversationalRetrievalChain.from_llm(llm=llm, retriever=vectorstore.as_retriever(), memory=memory)
def process_user_input(user_question):
response = st.session_state.conversation({'question': user_question})
st.session_state.chat_history = response['chat_history']
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)
filename = generate_filename(user_question, 'txt')
response = message.content
user_prompt = user_question
create_file(filename, user_prompt, response, should_save)
def divide_prompt(prompt, max_length):
words = prompt.split()
chunks = []
current_chunk = []
current_length = 0
for word in words:
if len(word) + current_length <= max_length:
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))
return chunks
# 13. Provide way of saving all and deleting all to give way of reviewing output and saving locally before clearing it
@st.cache_resource
def create_zip_of_files(files):
zip_name = "all_files.zip"
with zipfile.ZipFile(zip_name, 'w') as zipf:
for file in files:
zipf.write(file)
return zip_name
@st.cache_resource
def get_zip_download_link(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
# 14. Inference Endpoints for Whisper (best fastest STT) on NVIDIA T4 and Llama (best fastest AGI LLM) on NVIDIA A10
# My Inference Endpoint
API_URL_IE = f'https://tonpixzfvq3791u9.us-east-1.aws.endpoints.huggingface.cloud'
# Original
API_URL_IE = "https://api-inference.huggingface.co/models/openai/whisper-small.en"
MODEL2 = "openai/whisper-small.en"
MODEL2_URL = "https://huggingface.co/openai/whisper-small.en"
#headers = {
# "Authorization": "Bearer XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX",
# "Content-Type": "audio/wav"
#}
# HF_KEY = os.getenv('HF_KEY')
HF_KEY = st.secrets['HF_KEY']
headers = {
"Authorization": f"Bearer {HF_KEY}",
"Content-Type": "audio/wav"
}
#@st.cache_resource
def query(filename):
with open(filename, "rb") as f:
data = f.read()
response = requests.post(API_URL_IE, headers=headers, data=data)
return response.json()
def generate_filename(prompt, file_type):
central = pytz.timezone('US/Central')
safe_date_time = datetime.now(central).strftime("%m%d_%H%M")
replaced_prompt = prompt.replace(" ", "_").replace("\n", "_")
safe_prompt = "".join(x for x in replaced_prompt if x.isalnum() or x == "_")[:90]
return f"{safe_date_time}_{safe_prompt}.{file_type}"
# 15. Audio recorder to Wav file
def save_and_play_audio(audio_recorder):
audio_bytes = audio_recorder()
if audio_bytes:
filename = generate_filename("Recording", "wav")
with open(filename, 'wb') as f:
f.write(audio_bytes)
st.audio(audio_bytes, format="audio/wav")
return filename
# 16. Speech transcription to file output
def transcribe_audio(filename):
output = query(filename)
return output
def whisper_main():
#st.title("Speech to Text")
#st.write("Record your speech and get the text.")
# Audio, transcribe, GPT:
filename = save_and_play_audio(audio_recorder)
if filename is not None:
transcription = transcribe_audio(filename)
try:
transcript = transcription['text']
st.write(transcript)
except:
transcript=''
st.write(transcript)
# Whisper to GPT: New!! ---------------------------------------------------------------------
st.write('Reasoning with your inputs with GPT..')
response = chat_with_model(transcript)
st.write('Response:')
st.write(response)
filename = generate_filename(response, "txt")
create_file(filename, transcript, response, should_save)
# Whisper to GPT: New!! ---------------------------------------------------------------------
# Whisper to Llama:
#response = StreamLLMChatResponse(transcript)
#filename_txt = generate_filename(transcript, "md")
#create_file(filename_txt, transcript, response, should_save)
#filename_wav = filename_txt.replace('.txt', '.wav')
#import shutil
#try:
# if os.path.exists(filename):
# shutil.copyfile(filename, filename_wav)
#except:
# st.write('.')
#if os.path.exists(filename):
# os.remove(filename)
#st.experimental_rerun()
#except:
# st.write('Starting Whisper Model on GPU. Please retry in 30 seconds.')
# Sample function to demonstrate a response, replace with your own logic
def StreamMedChatResponse(topic):
st.write(f"Showing resources or questions related to: {topic}")
# 17. Main
def main():
prompt = f"Write ten funny jokes that are tweet length stories that make you laugh. Show as markdown outline with emojis for each."
# Add Wit and Humor buttons
# add_witty_humor_buttons()
# add_medical_exam_buttons()
with st.expander("Prompts 📚", expanded=False):
example_input = st.text_input("Enter your prompt text for Llama:", value=prompt, help="Enter text to get a response from DromeLlama.")
if st.button("Run Prompt With Llama model", help="Click to run the prompt."):
try:
response=StreamLLMChatResponse(example_input)
create_file(filename, example_input, response, should_save)
except:
st.write('Llama model is asleep. Starting now on A10 GPU. Please wait one minute then retry. KEDA triggered.')
openai.api_key = os.getenv('OPENAI_API_KEY')
if openai.api_key == None: openai.api_key = st.secrets['OPENAI_API_KEY']
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'))
user_prompt = st.text_area("Enter prompts, instructions & questions:", '', height=100)
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_sections = deque()
document_responses = {}
if uploaded_file is not None:
file_content = read_file_content(uploaded_file, max_length)
document_sections.extend(divide_document(file_content, max_length))
if len(document_sections) > 0:
if st.button("👁️ View Upload"):
st.markdown("**Sections of the uploaded file:**")
for i, section in enumerate(list(document_sections)):
st.markdown(f"**Section {i+1}**\n{section}")
st.markdown("**Chat with the model:**")
for i, section in enumerate(list(document_sections)):
if i in document_responses:
st.markdown(f"**Section {i+1}**\n{document_responses[i]}")
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)
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...')
user_prompt_sections = divide_prompt(user_prompt, max_length)
full_response = ''
for prompt_section in user_prompt_sections:
response = chat_with_model(prompt_section, ''.join(list(document_sections)), model_choice)
full_response += response + '\n' # Combine the responses
response = full_response
st.write('Response:')
st.write(response)
filename = generate_filename(user_prompt, choice)
create_file(filename, user_prompt, response, should_save)
# Compose a file sidebar of markdown md files:
all_files = glob.glob("*.md")
all_files = [file for file in all_files if len(os.path.splitext(file)[0]) >= 10] # 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
if st.sidebar.button("🗑 Delete All Text"):
for file in all_files:
os.remove(file)
st.experimental_rerun()
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)
file_contents=''
next_action=''
for file in all_files:
col1, col2, col3, col4, col5 = st.sidebar.columns([1,6,1,1,1]) # adjust the ratio as needed
with col1:
if st.button("🌐", key="md_"+file): # md emoji button
with open(file, 'r') as f:
file_contents = f.read()
next_action='md'
with col2:
st.markdown(get_table_download_link(file), unsafe_allow_html=True)
with col3:
if st.button("📂", key="open_"+file): # open emoji button
with open(file, 'r') as f:
file_contents = f.read()
next_action='open'
with col4:
if st.button("🔍", key="read_"+file): # search emoji button
with open(file, 'r') as f:
file_contents = f.read()
next_action='search'
with col5:
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)
if next_action=='md':
st.markdown(file_contents)
buttonlabel = '🔍Run with Llama and GPT.'
if st.button(key='RunWithLlamaandGPT', label = buttonlabel):
user_prompt = file_contents
# Llama versus GPT Battle!
all=""
try:
st.write('🔍Running with Llama.')
response = StreamLLMChatResponse(file_contents)
filename = generate_filename(user_prompt, "md")
create_file(filename, file_contents, response, should_save)
all=response
#SpeechSynthesis(response)
except:
st.markdown('Llama is sleeping. Restart ETA 30 seconds.')
# gpt
try:
st.write('🔍Running with GPT.')
response2 = chat_with_model(user_prompt, file_contents, model_choice)
filename2 = generate_filename(file_contents, choice)
create_file(filename2, user_prompt, response, should_save)
all=all+response2
#SpeechSynthesis(response2)
except:
st.markdown('GPT is sleeping. Restart ETA 30 seconds.')
SpeechSynthesis(all)
if next_action=='search':
file_content_area = st.text_area("File Contents:", file_contents, height=500)
st.write('🔍Running with Llama and GPT.')
user_prompt = file_contents
# Llama versus GPT Battle!
all=""
try:
st.write('🔍Running with Llama.')
response = StreamLLMChatResponse(file_contents)
filename = generate_filename(user_prompt, ".md")
create_file(filename, file_contents, response, should_save)
all=response
#SpeechSynthesis(response)
except:
st.markdown('Llama is sleeping. Restart ETA 30 seconds.')
# gpt
try:
st.write('🔍Running with GPT.')
response2 = chat_with_model(user_prompt, file_contents, model_choice)
filename2 = generate_filename(file_contents, choice)
create_file(filename2, user_prompt, response, should_save)
all=all+response2
#SpeechSynthesis(response2)
except:
st.markdown('GPT is sleeping. Restart ETA 30 seconds.')
SpeechSynthesis(all)
# Function to encode file to base64
def get_base64_encoded_file(file_path):
with open(file_path, "rb") as file:
return base64.b64encode(file.read()).decode()
# Function to create a download link
def get_audio_download_link(file_path):
base64_file = get_base64_encoded_file(file_path)
return f'<a href="data:file/wav;base64,{base64_file}" download="{os.path.basename(file_path)}">⬇️ Download Audio</a>'
# Compose a file sidebar of past encounters
all_files = glob.glob("*.wav")
all_files = [file for file in all_files if len(os.path.splitext(file)[0]) >= 10] # 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
filekey = 'delall'
if st.sidebar.button("🗑 Delete All Audio", key=filekey):
for file in all_files:
os.remove(file)
st.experimental_rerun()
for file in all_files:
col1, col2 = st.sidebar.columns([6, 1]) # adjust the ratio as needed
with col1:
st.markdown(file)
if st.button("🎵", key="play_" + file): # play emoji button
audio_file = open(file, 'rb')
audio_bytes = audio_file.read()
st.audio(audio_bytes, format='audio/wav')
#st.markdown(get_audio_download_link(file), unsafe_allow_html=True)
#st.text_input(label="", value=file)
with col2:
if st.button("🗑", key="delete_" + file):
os.remove(file)
st.experimental_rerun()
# Feedback
# Step: Give User a Way to Upvote or Downvote
GiveFeedback=False
if GiveFeedback:
with st.expander("Give your feedback 👍", expanded=False):
feedback = st.radio("Step 8: Give your feedback", ("👍 Upvote", "👎 Downvote"))
if feedback == "👍 Upvote":
st.write("You upvoted 👍. Thank you for your feedback!")
else:
st.write("You downvoted 👎. Thank you for your feedback!")
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)
with st.spinner("Processing"):
raw = pdf2txt(docs)
if len(raw) > 0:
length = str(len(raw))
text_chunks = txt2chunks(raw)
vectorstore = vector_store(text_chunks)
st.session_state.conversation = get_chain(vectorstore)
st.markdown('# AI Search Index of Length:' + length + ' Created.') # add timing
filename = generate_filename(raw, 'txt')
create_file(filename, raw, '', should_save)
# Relocated! Hope you like your new space - enjoy!
# Display instructions and handle query parameters
#st.markdown("## Glossary Lookup\nEnter a term in the URL query, like `?q=Nanotechnology` or `?query=Martian Syndicate`.")
st.markdown('''
### Mixable AI 🃏🚀📚
''')
try:
query_params = st.query_params
#query = (query_params.get('q') or query_params.get('query') or [''])[0]
query = (query_params.get('q') or query_params.get('query') or [''])
st.markdown('# Running query: ' + query)
if query: search_glossary(query)
except:
st.markdown('No glossary lookup')
# Display the glossary grid
st.title("Body Map Glossary 🎲")
display_glossary_grid(body_map_data)
display_buttons_with_scores()
display_images_and_wikipedia_summaries()
# Assuming the transhuman_glossary and other setup code remains the same
#st.write("Current Query Parameters:", st.query_params)
#st.markdown("### Query Parameters - These Deep Link Map to Remixable Methods, Navigate or Trigger Functionalities")
# Example: Using query parameters to navigate or trigger functionalities
if 'action' in st.query_params:
action = st.query_params()['action'][0] # Get the first (or only) 'action' parameter
if action == 'show_message':
st.success("Showing a message because 'action=show_message' was found in the URL.")
elif action == 'clear':
clear_query_params()
st.experimental_rerun()
# Handling repeated keys
if 'multi' in st.query_params:
multi_values = get_all_query_params('multi')
st.write("Values for 'multi':", multi_values)
# Manual entry for demonstration
st.write("Enter query parameters in the URL like this: ?action=show_message&multi=1&multi=2")
if 'query' in st.query_params:
query = st.query_params['query'][0] # Get the query parameter
# Display content or image based on the query
display_content_or_image(query)
# Add a clear query parameters button for convenience
if st.button("Clear Query Parameters", key='ClearQueryParams'):
# This will clear the browser URL's query parameters
st.experimental_set_query_params
st.experimental_rerun()
# 18. Run AI Pipeline
if __name__ == "__main__":
whisper_main()
main()