Spaces:
Running
on
CPU Upgrade
Running
on
CPU Upgrade
import streamlit as st | |
import streamlit.components.v1 as components | |
import os | |
import json | |
import random | |
import base64 | |
import glob | |
import math | |
import openai | |
import pytz | |
import re | |
import requests | |
import textract | |
import time | |
import zipfile | |
import dotenv | |
from gradio_client import Client | |
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 | |
from xml.etree import ElementTree as ET | |
from PIL import Image | |
from urllib.parse import quote # Ensure this import is included | |
# 1. Configuration | |
Site_Name = 'Scholarly-Article-Document-Search-With-Memory' | |
title="๐๐ArXiv Article Document Search Memory" | |
helpURL='https://huggingface.co/awacke1' | |
bugURL='https://huggingface.co/spaces/awacke1' | |
icons='๐๐๐๐' | |
st.set_page_config( | |
page_title=title, | |
page_icon=icons, | |
layout="wide", | |
initial_sidebar_state="expanded", | |
menu_items={ | |
'Get Help': helpURL, | |
'Report a bug': bugURL, | |
'About': title | |
} | |
) | |
def save_file(content, file_type): | |
timestamp = datetime.now().strftime("%Y%m%d_%H%M%S") | |
file_name = f"{file_type}_{timestamp}.md" | |
with open(file_name, "w") as file: | |
file.write(content) | |
return file_name | |
def load_file(file_name): | |
with open(file_name, "r", encoding='utf-8') as file: | |
#with open(file_name, "r") as file: | |
content = file.read() | |
return content | |
# HTML5 based Speech Synthesis (Text to Speech in Browser) | |
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>๐ Read It Aloud</h1> | |
<textarea id="textArea" rows="10" cols="80"> | |
''' | |
documentHTML5 = documentHTML5 + result | |
documentHTML5 = documentHTML5 + ''' | |
</textarea> | |
<br> | |
<button onclick="readAloud()">๐ Read Aloud</button> | |
</body> | |
</html> | |
''' | |
components.html(documentHTML5, width=1280, height=300) | |
def parse_to_markdown(text): | |
return text | |
def search_arxiv(query): | |
# Show ArXiv Scholary Articles! ----------------*************-------------***************---------------------------------------- | |
# st.title("โถ๏ธ Semantic and Episodic Memory System") | |
client = Client("awacke1/Arxiv-Paper-Search-And-QA-RAG-Pattern") | |
search_query = query | |
#top_n_results = st.slider(key='topnresults', label="Top n results as context", min_value=4, max_value=100, value=100) | |
#search_source = st.sidebar.selectbox(key='searchsource', label="Search Source", ["Semantic Search - up to 10 Mar 2024", "Arxiv Search - Latest - (EXPERIMENTAL)"]) | |
search_source = "Arxiv Search - Latest - (EXPERIMENTAL)" # "Semantic Search - up to 10 Mar 2024" | |
#llm_model = st.sidebar.selectbox(key='llmmodel', label="LLM Model", ["mistralai/Mixtral-8x7B-Instruct-v0.1", "mistralai/Mistral-7B-Instruct-v0.2", "google/gemma-7b-it", "None"]) | |
llm_model = "mistralai/Mixtral-8x7B-Instruct-v0.1" | |
st.sidebar.markdown('### ๐ ' + query) | |
# ArXiv searcher ~-<>-~ Paper Summary - Ask LLM | |
client = Client("awacke1/Arxiv-Paper-Search-And-QA-RAG-Pattern") | |
response2 = client.predict( | |
query, # str in 'parameter_13' Textbox component | |
"mistralai/Mixtral-8x7B-Instruct-v0.1", # Literal['mistralai/Mixtral-8x7B-Instruct-v0.1', 'mistralai/Mistral-7B-Instruct-v0.2', 'google/gemma-7b-it', 'None'] in 'LLM Model' Dropdown component | |
True, # bool in 'Stream output' Checkbox component | |
api_name="/ask_llm" | |
) | |
st.write('๐Run of Multi-Agent System Paper Summary Spec is Complete') | |
st.markdown(response2) | |
# ArXiv searcher ~-<>-~ Paper References - Update with RAG | |
client = Client("awacke1/Arxiv-Paper-Search-And-QA-RAG-Pattern") | |
response1 = client.predict( | |
query, | |
10, | |
"Semantic Search - up to 10 Mar 2024", # Literal['Semantic Search - up to 10 Mar 2024', 'Arxiv Search - Latest - (EXPERIMENTAL)'] in 'Search Source' Dropdown component | |
"mistralai/Mixtral-8x7B-Instruct-v0.1", # Literal['mistralai/Mixtral-8x7B-Instruct-v0.1', 'mistralai/Mistral-7B-Instruct-v0.2', 'google/gemma-7b-it', 'None'] in 'LLM Model' Dropdown component | |
api_name="/update_with_rag_md" | |
) | |
st.write('๐Run of Multi-Agent System Paper References is Complete') | |
responseall = response1[0] + response1[1] | |
st.markdown(responseall) | |
result = response2 + responseall | |
SpeechSynthesis(result) # Search History Reader / Writer IO Memory - Audio at Same time as Reading. | |
filename=generate_filename(query, "md") | |
create_file(filename, query, result, should_save) | |
saved_files = [f for f in os.listdir(".") if f.endswith(".md")] | |
selected_file = st.sidebar.selectbox("Saved Files", saved_files) | |
if selected_file: | |
file_content = load_file(selected_file) | |
st.sidebar.markdown(file_content) | |
if st.sidebar.button("๐๏ธ Delete"): | |
os.remove(selected_file) | |
st.warning(f"File deleted: {selected_file}") | |
return result | |
# Prompts for App, for App Product, and App Product Code | |
PromptPrefix = 'Create a specification with streamlit functions creating markdown outlines and tables rich with appropriate emojis for methodical step by step rules defining the concepts at play. Use story structure architect rules to plan, structure and write three dramatic situations to include in the rules and how to play by matching the theme for topic of ' | |
PromptPrefix2 = 'Create a streamlit python user app with full code listing to create a UI implementing the using streamlit, gradio, huggingface to create user interface elements like emoji buttons, sliders, drop downs, and data interfaces like dataframes to show tables, session_statematching this ruleset and thematic story plot line: ' | |
PromptPrefix3 = 'Create a HTML5 aframe and javascript app using appropriate libraries to create a word game simulation with advanced libraries like aframe to render 3d scenes creating moving entities that stay within a bounding box but show text and animation in 3d for inventory, components and story entities. Show full code listing. Add a list of new random entities say 3 of a few different types to any list appropriately and use emojis to make things easier and fun to read. Use appropriate emojis in labels. Create the UI to implement storytelling in the style of a dungeon master, with features using three emoji appropriate text plot twists and recurring interesting funny fascinating and complex almost poetic named characters with genius traits and file IO, randomness, ten point choice lists, math distribution tradeoffs, witty humorous dilemnas with emoji , rewards, variables, reusable functions with parameters, and data driven app with python libraries and streamlit components for Javascript and HTML5. Use appropriate emojis for labels to summarize and list parts, function, conditions for topic:' | |
# Function to display the entire glossary in a grid format with links | |
def display_glossary_grid(roleplaying_glossary): | |
search_urls = { | |
"๐Wiki": lambda k: f"https://en.wikipedia.org/wiki/{quote(k)}", | |
"๐Google": lambda k: f"https://www.google.com/search?q={quote(k)}", | |
"โถ๏ธYouTube": lambda k: f"https://www.youtube.com/results?search_query={quote(k)}", | |
"๐Bing": lambda k: f"https://www.bing.com/search?q={quote(k)}", | |
"๐ฅYouTube": lambda k: f"https://www.youtube.com/results?search_query={quote(k)}", | |
"๐ฆTwitter": lambda k: f"https://twitter.com/search?q={quote(k)}", | |
"๐๐ArXiv": lambda k: f"https://huggingface.co/spaces/awacke1/{Site_Name}?q={quote(k)}", # this url plus query! | |
"๐Analyst": lambda k: f"https://huggingface.co/spaces/awacke1/{Site_Name}?q={quote(k)}-{quote(PromptPrefix)}", # this url plus query! | |
"๐PyCoder": lambda k: f"https://huggingface.co/spaces/awacke1/{Site_Name}?q={quote(k)}-{quote(PromptPrefix2)}", # this url plus query! | |
"๐ฌJSCoder": lambda k: f"https://huggingface.co/spaces/awacke1/{Site_Name}?q={quote(k)}-{quote(PromptPrefix3)}", # this url plus query! | |
} | |
for category, details in roleplaying_glossary.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: | |
gameterm = category + ' - ' + game + ' - ' + term | |
links_md = ' '.join([f"[{emoji}]({url(gameterm)})" for emoji, url in search_urls.items()]) | |
#links_md = ' '.join([f"[{emoji}]({url(term)})" for emoji, url in search_urls.items()]) | |
#st.markdown(f"{term} {links_md}", unsafe_allow_html=True) | |
st.markdown(f"**{term}** <small>{links_md}</small>", unsafe_allow_html=True) | |
def display_glossary_entity(k): | |
search_urls = { | |
"๐Wiki": lambda k: f"https://en.wikipedia.org/wiki/{quote(k)}", | |
"๐Google": lambda k: f"https://www.google.com/search?q={quote(k)}", | |
"โถ๏ธYouTube": lambda k: f"https://www.youtube.com/results?search_query={quote(k)}", | |
"๐Bing": lambda k: f"https://www.bing.com/search?q={quote(k)}", | |
"๐ฅYouTube": lambda k: f"https://www.youtube.com/results?search_query={quote(k)}", | |
"๐ฆTwitter": lambda k: f"https://twitter.com/search?q={quote(k)}", | |
"๐๐ArXiv": lambda k: f"https://huggingface.co/spaces/awacke1/{Site_Name}?q={quote(k)}", # this url plus query! | |
"๐Analyst": lambda k: f"https://huggingface.co/spaces/awacke1/{Site_Name}?q={quote(k)}-{quote(PromptPrefix)}", # this url plus query! | |
"๐PyCoder": lambda k: f"https://huggingface.co/spaces/awacke1/{Site_Name}?q={quote(k)}-{quote(PromptPrefix2)}", # this url plus query! | |
"๐ฌJSCoder": lambda k: f"https://huggingface.co/spaces/awacke1/{Site_Name}?q={quote(k)}-{quote(PromptPrefix3)}", # this url plus query! | |
} | |
links_md = ' '.join([f"[{emoji}]({url(k)})" for emoji, url in search_urls.items()]) | |
#st.markdown(f"{k} {links_md}", unsafe_allow_html=True) | |
st.markdown(f"**{k}** <small>{links_md}</small>", unsafe_allow_html=True) | |
roleplaying_glossary = { | |
"๐ค AI Concepts": { | |
"MoE (Mixture of Experts) ๐ง ": [ | |
"What are Multi Agent Systems for Health", | |
"What is Mixture of Experts for Health", | |
"What are Semantic and Episodic Memory and what is Mirroring for Behavioral Health", | |
"What are Self Rewarding AI Systems for Health", | |
"How are AGI and AMI systems created using Multi Agent Systems and Mixture of Experts for Health" | |
], | |
"Multi Agent Systems (MAS) ๐ค": [ | |
"Distributed AI systems", | |
"Autonomous agents interacting", | |
"Cooperative and competitive behavior", | |
"Decentralized problem-solving", | |
"Applications in robotics, simulations, and more" | |
], | |
"Self Rewarding AI ๐": [ | |
"Intrinsic motivation for AI agents", | |
"Autonomous goal setting and achievement", | |
"Exploration and curiosity-driven learning", | |
"Potential for open-ended development", | |
"Research area in reinforcement learning" | |
], | |
"Semantic and Episodic Memory ๐": [ | |
"Two types of long-term memory", | |
"Semantic: facts and general knowledge", | |
"Episodic: personal experiences and events", | |
"Crucial for AI systems to understand and reason", | |
"Research in knowledge representation and retrieval" | |
] | |
}, | |
"๐ ๏ธ AI Tools & Platforms": { | |
"AutoGen ๐ง": [ | |
"Automated machine learning (AutoML) tool", | |
"Generates AI models based on requirements", | |
"Simplifies AI development process", | |
"Accessible to non-experts", | |
"Integration with various data sources" | |
], | |
"ChatDev ๐ฌ": [ | |
"Platform for building chatbots and conversational AI", | |
"Drag-and-drop interface for designing chat flows", | |
"Pre-built templates and integrations", | |
"Supports multiple messaging platforms", | |
"Analytics and performance tracking" | |
], | |
"Omniverse ๐": [ | |
"Nvidia's 3D simulation and collaboration platform", | |
"Physically accurate virtual worlds", | |
"Supports AI training and testing", | |
"Used in industries like robotics, architecture, and gaming", | |
"Enables seamless collaboration and data exchange" | |
], | |
"Lumiere ๐ฅ": [ | |
"AI-powered video analytics platform", | |
"Extracts insights and metadata from video content", | |
"Facial recognition and object detection", | |
"Sentiment analysis and scene understanding", | |
"Applications in security, media, and marketing" | |
], | |
"SORA ๐๏ธ": [ | |
"Scalable Open Research Architecture", | |
"Framework for distributed AI research and development", | |
"Modular and extensible design", | |
"Facilitates collaboration and reproducibility", | |
"Supports various AI algorithms and models" | |
] | |
}, | |
"๐ฌ Science Topics": { | |
"Physics ๐ญ": [ | |
"Astrophysics: galaxies, cosmology, planets, high energy phenomena, instrumentation, solar/stellar", | |
"Condensed Matter: disordered systems, materials science, nano/mesoscale, quantum gases, soft matter, statistical mechanics, superconductivity", | |
"General Relativity and Quantum Cosmology", | |
"High Energy Physics: experiment, lattice, phenomenology, theory", | |
"Mathematical Physics", | |
"Nonlinear Sciences: adaptation, cellular automata, chaos, solvable systems, pattern formation", | |
"Nuclear: experiment, theory", | |
"Physics: accelerators, atmospherics, atomic/molecular, biophysics, chemical, computational, education, fluids, geophysics, optics, plasma, popular, space" | |
], | |
"Mathematics โ": [ | |
"Algebra: geometry, topology, number theory, combinatorics, representation theory", | |
"Analysis: PDEs, functional, numerical, spectral theory, ODEs, complex variables", | |
"Geometry: algebraic, differential, metric, symplectic, topological", | |
"Probability and Statistics", | |
"Applied Math: information theory, optimization and control" | |
], | |
"Computer Science ๐ป": [ | |
"Artificial Intelligence and Machine Learning", | |
"Computation and Language, Complexity, Engineering, Finance, Science", | |
"Computer Vision, Graphics, Robotics", | |
"Cryptography, Security, Blockchain", | |
"Data Structures, Algorithms, Databases", | |
"Distributed and Parallel Computing", | |
"Formal Languages, Automata, Logic", | |
"Information Theory, Signal Processing", | |
"Networks, Internet Architecture, Social Networks", | |
"Programming Languages, Software Engineering" | |
], | |
"Quantitative Biology ๐งฌ": [ | |
"Biomolecules, Cell Behavior, Genomics", | |
"Molecular Networks, Neurons and Cognition", | |
"Populations, Evolution, Ecology", | |
"Quantitative Methods, Subcellular Processes", | |
"Tissues, Organs, Organisms" | |
], | |
"Quantitative Finance ๐": [ | |
"Computational and Mathematical Finance", | |
"Econometrics and Statistical Finance", | |
"Economics, Portfolio Management, Trading", | |
"Pricing, Risk Management" | |
], | |
"Electrical Engineering ๐": [ | |
"Audio, Speech, Image and Video Processing", | |
"Communications and Information Theory", | |
"Signal Processing, Controls, Robotics", | |
"Electronic Circuits, Embedded Systems" | |
] | |
} | |
} | |
def get_table_download_link(file_path): | |
#with open(file_path, 'r') as file: | |
#with open(file_path, 'r', encoding="unicode", errors="surrogateescape") as file: | |
with open(file_path, 'r', encoding='utf-8') 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 create_zip_of_files(files): # ---------------------------------- | |
zip_name = "Arxiv-Paper-Search-QA-RAG-Streamlit-Gradio-AP.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): | |
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 get_file(): | |
st.write(st.session_state['file']) | |
def SaveFileTextClicked(): | |
fileText = st.session_state.file_content_area | |
fileName = st.session_state.file_name_input | |
with open(fileName, 'w', encoding='utf-8') as file: | |
file.write(fileText) | |
st.markdown('Saved ' + fileName + '.') | |
def SaveFileNameClicked(): | |
newFileName = st.session_state.file_name_input | |
oldFileName = st.session_state.filename | |
if (newFileName!=oldFileName): | |
os.rename(oldFileName, newFileName) | |
st.markdown('Renamed file ' + oldFileName + ' to ' + newFileName + '.') | |
newFileText = st.session_state.file_content_area | |
oldFileText = st.session_state.filetext | |
def FileSidebar(): | |
# ----------------------------------------------------- File Sidebar for Jump Gates ------------------------------------------ | |
# 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 | |
# Delete and Download: | |
Files1, Files2 = st.sidebar.columns(2) | |
with Files1: | |
if st.button("๐ Delete All"): | |
for file in all_files: | |
os.remove(file) | |
st.experimental_rerun() | |
with Files2: | |
if st.button("โฌ๏ธ Download"): | |
zip_file = create_zip_of_files(all_files) | |
st.sidebar.markdown(get_zip_download_link(zip_file), unsafe_allow_html=True) | |
file_contents='' | |
file_name='' | |
next_action='' | |
# FileSidebar Adds files matching pattern to sidebar where we feature a ๐View, ๐Open, ๐Run, and ๐Delete per file | |
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("๐", key="md_"+file): # md emoji button | |
file_contents = load_file(file) | |
file_name=file | |
next_action='md' | |
st.session_state['next_action'] = next_action | |
with col2: | |
st.markdown(get_table_download_link(file), unsafe_allow_html=True) | |
with col3: | |
if st.button("๐", key="open_"+file): # open emoji button | |
file_contents = load_file(file) | |
file_name=file | |
next_action='open' | |
st.session_state['lastfilename'] = file | |
st.session_state['filename'] = file | |
st.session_state['filetext'] = file_contents | |
st.session_state['next_action'] = next_action | |
with col4: | |
if st.button("๐", key="read_"+file): # search emoji button | |
file_contents = load_file(file) | |
file_name=file | |
next_action='search' | |
st.session_state['next_action'] = next_action | |
with col5: | |
if st.button("๐", key="delete_"+file): | |
os.remove(file) | |
file_name=file | |
st.experimental_rerun() | |
next_action='delete' | |
st.session_state['next_action'] = next_action | |
if len(file_contents) > 0: | |
if next_action=='open': # For "open", prep session state if it hasn't been yet | |
if 'lastfilename' not in st.session_state: | |
st.session_state['lastfilename'] = '' | |
if 'filename' not in st.session_state: | |
st.session_state['filename'] = '' | |
if 'filetext' not in st.session_state: | |
st.session_state['filetext'] = '' | |
open1, open2 = st.columns(spec=[.8,.2]) | |
with open1: | |
# Use onchange functions to autoexecute file name and text save functions. | |
file_name_input = st.text_input(key='file_name_input', on_change=SaveFileNameClicked, label="File Name:",value=file_name ) | |
file_content_area = st.text_area(key='file_content_area', on_change=SaveFileTextClicked, label="File Contents:", value=file_contents, height=300) | |
bp1,bp2 = st.columns([.5,.5]) | |
with bp1: | |
if st.button(label='๐พ Save Name'): | |
SaveFileNameClicked() | |
with bp2: | |
if st.button(label='๐พ Save File'): | |
SaveFileTextClicked() | |
new_file_content_area = st.session_state['file_content_area'] | |
if new_file_content_area != file_contents: | |
st.markdown(new_file_content_area) #changed | |
if st.button("๐ Run AI Meta Strategy", key="filecontentssearch"): | |
#search_glossary(file_content_area) | |
filesearch = PromptPrefix + file_content_area | |
st.markdown(filesearch) | |
if st.button(key=rerun, label='๐Re-Spec' ): | |
search_glossary(filesearch) | |
if next_action=='md': | |
st.markdown(file_contents) | |
buttonlabel = '๐Run' | |
if st.button(key='Runmd', label = buttonlabel): | |
user_prompt = file_contents | |
#try: | |
search_glossary(file_contents) | |
#except: | |
#st.markdown('GPT is sleeping. Restart ETA 30 seconds.') | |
if next_action=='search': | |
file_content_area = st.text_area("File Contents:", file_contents, height=500) | |
user_prompt = file_contents | |
#try: | |
#search_glossary(file_contents) | |
filesearch = PromptPrefix2 + file_content_area | |
st.markdown(filesearch) | |
if st.button(key=rerun, label='๐Re-Code' ): | |
search_glossary(filesearch) | |
#except: | |
#st.markdown('GPT is sleeping. Restart ETA 30 seconds.') | |
# ----------------------------------------------------- File Sidebar for Jump Gates ------------------------------------------ | |
st.markdown("### ๐ฒ๐บ๏ธ Scholarly Article Document Search Memory") | |
FileSidebar() | |
# ---- Art Card Sidebar with Random Selection of image: | |
def get_image_as_base64(url): | |
response = requests.get(url) | |
if response.status_code == 200: | |
# Convert the image to base64 | |
return base64.b64encode(response.content).decode("utf-8") | |
else: | |
return None | |
def create_download_link(filename, base64_str): | |
href = f'<a href="data:file/png;base64,{base64_str}" download="{filename}">Download Image</a>' | |
return href | |
def SideBarImageShuffle(): | |
image_urls = [ | |
"https://cdn-uploads.huggingface.co/production/uploads/620630b603825909dcbeba35/cfhJIasuxLkT5fnaAE6Gj.png", | |
"https://cdn-uploads.huggingface.co/production/uploads/620630b603825909dcbeba35/UMo4oWNrrd6RLLzsFxQAi.png", | |
"https://cdn-uploads.huggingface.co/production/uploads/620630b603825909dcbeba35/o_EH4cTs5Qxiu7xTZw9I3.png", | |
"https://cdn-uploads.huggingface.co/production/uploads/620630b603825909dcbeba35/cmCZ5RTdSx3usMm7MwwWK.png", | |
] | |
selected_image_url = random.choice(image_urls) | |
selected_image_base64 = get_image_as_base64(selected_image_url) | |
if selected_image_base64 is not None: | |
with st.sidebar: | |
st.markdown(f"![image](data:image/png;base64,{selected_image_base64})") | |
else: | |
st.sidebar.write("Failed to load the image.") | |
ShowSideImages=False | |
if ShowSideImages: | |
SideBarImageShuffle() | |
# 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 | |
# ๐Run-------------------------------------------------------- | |
def search_glossary(query): | |
for category, terms in roleplaying_glossary.items(): | |
if query.lower() in (term.lower() for term in terms): | |
st.markdown(f"#### {category}") | |
st.write(f"- {query}") | |
all="" | |
# ๐Run 1 - plain query | |
#response = chat_with_model(query) | |
#response1 = chat_with_model45(query) | |
#all = query + ' ' + response1 | |
#st.write('๐Run 1 is Complete.') | |
# ArXiv searcher ~-<>-~ Paper Summary - Ask LLM | |
client = Client("awacke1/Arxiv-Paper-Search-And-QA-RAG-Pattern") | |
response2 = client.predict( | |
query, # str in 'parameter_13' Textbox component | |
"mistralai/Mixtral-8x7B-Instruct-v0.1", # Literal['mistralai/Mixtral-8x7B-Instruct-v0.1', 'mistralai/Mistral-7B-Instruct-v0.2', 'google/gemma-7b-it', 'None'] in 'LLM Model' Dropdown component | |
True, # bool in 'Stream output' Checkbox component | |
api_name="/ask_llm" | |
) | |
st.write('๐Run of Multi-Agent System Paper Summary Spec is Complete') | |
#st.markdown(response2) | |
# ArXiv searcher ~-<>-~ Paper References - Update with RAG | |
client = Client("awacke1/Arxiv-Paper-Search-And-QA-RAG-Pattern") | |
response1 = client.predict( | |
query, | |
10, | |
"Semantic Search - up to 10 Mar 2024", # Literal['Semantic Search - up to 10 Mar 2024', 'Arxiv Search - Latest - (EXPERIMENTAL)'] in 'Search Source' Dropdown component | |
"mistralai/Mixtral-8x7B-Instruct-v0.1", # Literal['mistralai/Mixtral-8x7B-Instruct-v0.1', 'mistralai/Mistral-7B-Instruct-v0.2', 'google/gemma-7b-it', 'None'] in 'LLM Model' Dropdown component | |
api_name="/update_with_rag_md" | |
) | |
st.write('๐Run of Multi-Agent System Paper References is Complete') | |
#st.markdown(response1) | |
responseall = response2 + response1[0] + response1[1] | |
st.markdown(responseall) | |
return responseall | |
# GPT 35 turbo and GPT 45 - - - - - - - - - - - - -<><><><><>: | |
RunPostArxivLLM = False | |
if RunPostArxivLLM: | |
# ๐Run PaperSummarizer | |
PaperSummarizer = ' Create a paper summary as a markdown table with paper links clustering the features writing short markdown emoji outlines to extract three main ideas from each of the ten summaries. For each one create three simple points led by an emoji of the main three steps needed as method step process for implementing the idea as a single app.py streamlit python app. ' | |
response2 = chat_with_model(PaperSummarizer + str(response1)) | |
st.write('๐Run 3 - Paper Summarizer is Complete.') | |
# ๐Run AppSpecifier | |
AppSpecifier = ' Design and write a streamlit python code listing and specification that implements each scientific method steps as ten functions keeping specification in a markdown table in the function comments with original paper link to outline the AI pipeline ensemble implementing code as full plan to build.' | |
response3 = chat_with_model(AppSpecifier + str(response2)) | |
st.write('๐Run 4 - AppSpecifier is Complete.') | |
# ๐Run PythonAppCoder | |
PythonAppCoder = ' Complete this streamlit python app implementing the functions in detail using appropriate python libraries and streamlit user interface elements. Show full code listing for the completed detail app as full code listing with no comments or commentary. ' | |
#result = str(result).replace('\n', ' ').replace('|', ' ') | |
# response4 = chat_with_model45(PythonAppCoder + str(response3)) | |
response4 = chat_with_model(PythonAppCoder + str(response3)) | |
st.write('๐Run Python AppCoder is Complete.') | |
# experimental 45 - - - - - - - - - - - - -<><><><><> | |
responseAll = '# Query: ' + query + '# Summary: ' + str(response2) + '# Streamlit App Specifier: ' + str(response3) + '# Complete Streamlit App: ' + str(response4) + '# Scholarly Article Links References: ' + str(response1) | |
filename = generate_filename(responseAll, "md") | |
create_file(filename, query, responseAll, should_save) | |
return responseAll # ๐Run-------------------------------------------------------- | |
else: | |
return response1 | |
# Function to display the glossary in a structured format | |
def display_glossary(glossary, area): | |
if area in glossary: | |
st.subheader(f"๐ 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(roleplaying_glossary): | |
search_urls = { | |
"๐Wiki": lambda k: f"https://en.wikipedia.org/wiki/{quote(k)}", | |
"๐Google": lambda k: f"https://www.google.com/search?q={quote(k)}", | |
"โถ๏ธYouTube": lambda k: f"https://www.youtube.com/results?search_query={quote(k)}", | |
"๐Bing": lambda k: f"https://www.bing.com/search?q={quote(k)}", | |
"๐ฅYouTube": lambda k: f"https://www.youtube.com/results?search_query={quote(k)}", | |
"๐ฆTwitter": lambda k: f"https://twitter.com/search?q={quote(k)}", | |
"๐๐ArXiv": lambda k: f"https://huggingface.co/spaces/awacke1/{Site_Name}?q={quote(k)}", # this url plus query! | |
"๐Analyst": lambda k: f"https://huggingface.co/spaces/awacke1/{Site_Name}?q={quote(k)}-{quote(PromptPrefix)}", # this url plus query! | |
"๐PyCoder": lambda k: f"https://huggingface.co/spaces/awacke1/{Site_Name}?q={quote(k)}-{quote(PromptPrefix2)}", # this url plus query! | |
"๐ฌJSCoder": lambda k: f"https://huggingface.co/spaces/awacke1/{Site_Name}?q={quote(k)}-{quote(PromptPrefix3)}", # this url plus query! | |
} | |
for category, details in roleplaying_glossary.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}** <small>{links_md}</small>", unsafe_allow_html=True) | |
def display_videos_and_links(): | |
video_files = [f for f in os.listdir('.') if f.endswith('.mp4')] | |
if not video_files: | |
st.write("No MP4 videos found in the current directory.") | |
return | |
video_files_sorted = sorted(video_files, key=lambda x: len(x.split('.')[0])) | |
num_columns=4 | |
cols = st.columns(num_columns) # Define num_columns columns outside the loop | |
col_index = 0 # Initialize column index | |
for video_file in video_files_sorted: | |
with cols[col_index % num_columns]: # Use modulo 2 to alternate between the first and second column | |
# Embedding video with autoplay and loop using HTML | |
#video_html = ("""<video width="100%" loop autoplay> <source src="{video_file}" type="video/mp4">Your browser does not support the video tag.</video>""") | |
#st.markdown(video_html, unsafe_allow_html=True) | |
k = video_file.split('.')[0] # Assumes keyword is the file name without extension | |
st.video(video_file, format='video/mp4', start_time=0) | |
display_glossary_entity(k) | |
col_index += 1 # Increment column index to place the next video in the next column | |
#@st.cache_resource | |
def display_images_and_wikipedia_summaries(num_columns=4): | |
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 | |
image_files_sorted = sorted(image_files, key=lambda x: len(x.split('.')[0])) | |
cols = st.columns(num_columns) # Use specified num_columns for layout | |
col_index = 0 # Initialize column index for cycling through columns | |
for image_file in image_files_sorted: | |
with cols[col_index % num_columns]: # Cycle through columns based on num_columns | |
image = Image.open(image_file) | |
st.image(image, caption=image_file, use_column_width=True) | |
k = image_file.split('.')[0] # Assumes keyword is the file name without extension | |
display_glossary_entity(k) | |
col_index += 1 # Increment to move to the next column in the next iteration | |
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): | |
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 | |
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 | |
st.warning("No matching content or image found.") | |
return False | |
game_emojis = { | |
"Dungeons and Dragons": "๐", | |
"Call of Cthulhu": "๐", | |
"GURPS": "๐ฒ", | |
"Pathfinder": "๐บ๏ธ", | |
"Kindred of the East": "๐ ", | |
"Changeling": "๐", | |
} | |
topic_emojis = { | |
"Core Rulebooks": "๐", | |
"Maps & Settings": "๐บ๏ธ", | |
"Game Mechanics & Tools": "โ๏ธ", | |
"Monsters & Adversaries": "๐น", | |
"Campaigns & Adventures": "๐", | |
"Creatives & Assets": "๐จ", | |
"Game Master Resources": "๐ ๏ธ", | |
"Lore & Background": "๐", | |
"Character Development": "๐ง", | |
"Homebrew Content": "๐ง", | |
"General Topics": "๐", | |
} | |
# Adjusted display_buttons_with_scores function | |
def display_buttons_with_scores(): | |
for category, games in roleplaying_glossary.items(): | |
category_emoji = topic_emojis.get(category, "๐") # 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, "๐ฎ") # 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} {category} {game} {term} {score}", key=key): | |
update_score(key) | |
# Create a dynamic query incorporating emojis and formatting for clarity | |
query_prefix = f"{category_emoji} {game_emoji} ** {category} - {game} - {term} - **" | |
# ---------------------------------------------------------------------------------------------- | |
#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 emoji laden user interface with labels with the entity name and emojis in all labels with a set of streamlit UI components with drop down lists and dataframes and buttons with expander and sidebar for the app to run the data as default values mostly in text boxes. Feature a 3 point outline sith 3 subpoints each where each line has about six words describing this and also contain appropriate emoji for creating sumamry of all aspeccts of this topic. 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) | |
def get_all_query_params(key): | |
return st.query_params().get(key, []) | |
def clear_query_params(): | |
st.query_params() | |
# 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 = "...." | |
should_save = st.sidebar.checkbox("๐พ Save", value=True, help="Save your session data.") | |
# 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']: | |
# ****** line 344 is read utf-8 encoding was needed when running locally to save utf-8 encoding and not fail on write | |
#with open(f"{base_filename}.md", 'w') as file: | |
#with open(f"{base_filename}.md", 'w', encoding="ascii", errors="surrogateescape") as file: | |
with open(f"{base_filename}.md", 'w', encoding='utf-8') as file: | |
#try: | |
#content = (prompt.strip() + '\r\n' + decode(response, )) | |
file.write(response) | |
#except: | |
# st.write('.') | |
# ****** utf-8 encoding was needed when running locally to save utf-8 encoding and not fail on 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)] | |
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 | |
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 - now favoring fastest AI pipeline STT Whisper->LLM Llama->TTS | |
def chat_with_model(prompt, document_section='', model_choice='gpt-3.5-turbo'): # gpt-4-0125-preview gpt-3.5-turbo | |
model = model_choice | |
conversation = [{'role': 'system', 'content': 'You are a coder, inventor, and writer of quotes on wisdom as a helpful expert in all fields of health, math, development and AI using python.'}] | |
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=model_choice, 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 | |
# 11.1 45 | |
def chat_with_model45(prompt, document_section='', model_choice='gpt-4-0125-preview'): # gpt-4-0125-preview gpt-3.5-turbo | |
model = model_choice | |
conversation = [{'role': 'system', 'content': 'You are a coder, inventor, and writer of quotes on wisdom as a helpful expert in all fields of health, math, development and AI using python.'}] | |
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=model_choice, 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 chat_with_file_contents(prompt, file_content, model_choice='gpt-3.5-turbo'): # gpt-4-0125-preview gpt-3.5-turbo | |
#def chat_with_file_contents(prompt, file_content, model_choice='gpt-4-0125-preview'): # gpt-4-0125-preview gpt-3.5-turbo | |
conversation = [{'role': 'system', 'content': 'You are a helpful assistant.'}] | |
conversation.append({'role': 'user', 'content': prompt}) | |
if len(file_content)>0: | |
conversation.append({'role': 'assistant', 'content': file_content}) | |
response = openai.ChatCompletion.create(model=model_choice, messages=conversation) | |
return response['choices'][0]['message']['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 | |
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 | |
def vector_store(text_chunks): | |
embeddings = OpenAIEmbeddings(openai_api_key=key) | |
return FAISS.from_texts(texts=text_chunks, embedding=embeddings) | |
# Memory and Retrieval chains | |
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 | |
API_URL_IE = f'https://tonpixzfvq3791u9.us-east-1.aws.endpoints.huggingface.cloud' | |
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" | |
HF_KEY = st.secrets['HF_KEY'] | |
headers = { | |
"Authorization": f"Bearer {HF_KEY}", | |
"Content-Type": "audio/wav" | |
} | |
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 | |
# Sample function to demonstrate a response, replace with your own logic | |
def StreamMedChatResponse(topic): | |
st.write(f"Showing resources or questions related to: {topic}") | |
# 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)}">โฌ๏ธ Download Audio</a>' | |
# Wav audio files - Transcription History in Wav | |
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("๐ 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("๐ต", 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("๐", key="delete_" + file): | |
os.remove(file) | |
st.experimental_rerun() | |
GiveFeedback=False | |
if GiveFeedback: | |
with st.expander("Give your feedback ๐", expanded=False): | |
feedback = st.radio("Step 8: Give your feedback", ("๐ Upvote", "๐ Downvote")) | |
if feedback == "๐ Upvote": | |
st.write("You upvoted ๐. Thank you for your feedback!") | |
else: | |
st.write("You downvoted ๐. 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) | |
try: | |
query_params = st.query_params | |
query = (query_params.get('q') or query_params.get('query') or ['']) | |
if query: | |
result = search_arxiv(query) | |
#result2 = search_glossary(result) | |
except: | |
st.markdown(' ') | |
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() | |
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) | |
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) | |
st.write('Reasoning with your inputs..') | |
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 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) | |
prompt = ''' | |
What is MoE? | |
What are Multi Agent Systems? | |
What is Self Rewarding AI? | |
What is Semantic and Episodic memory? | |
What is AutoGen? | |
What is ChatDev? | |
What is Omniverse? | |
What is Lumiere? | |
What is SORA? | |
''' | |
session_state = {} | |
if "search_queries" not in session_state: | |
session_state["search_queries"] = [] | |
example_input = st.text_input("Search", value=session_state["search_queries"][-1] if session_state["search_queries"] else "") | |
if example_input: | |
session_state["search_queries"].append(example_input) | |
# Search AI | |
query=example_input | |
if query: | |
result = search_arxiv(query) | |
#search_glossary(query) | |
search_glossary(result) | |
st.markdown(' ') | |
st.write("Search history:") | |
for example_input in session_state["search_queries"]: | |
st.write(example_input) | |
if st.button("Run Prompt", help="Click to run."): | |
try: | |
response=StreamLLMChatResponse(example_input) | |
create_file(filename, example_input, response, should_save) | |
except: | |
st.write('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(key='maxlength', label="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("๐๏ธ 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...') | |
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) | |
display_images_and_wikipedia_summaries() # Image Jump Grid | |
display_videos_and_links() # Video Jump Grid | |
display_glossary_grid(roleplaying_glossary) # Word Glossary Jump Grid | |
display_buttons_with_scores() # Feedback Jump Grid | |