import os
import ssl
import gradio as gr
from dotenv import load_dotenv
from config.config import TALKING_HEAD_WIDTH, TOOLS_DEFAULT_LIST, NUM_WORDS_DEFAULT, FORMALITY_DEFAULT, EMOTION_DEFAULT, \
LANG_LEVEL_DEFAULT, TRANSLATE_TO_DEFAULT, LITERARY_STYLE_DEFAULT, LOOPING_TALKING_HEAD_VIDEO_PATH, TOOLS_LIST, \
MAX_WORDS
from utilities.audio import FORCE_TRANSLATE_DEFAULT, USE_GPT4_DEFAULT, transcribe, WHISPER_DETECT_LANG
from utilities.html_stuff import create_html_video, update_talking_head
from utilities.reset_memory import ChatWrapper, reset_memory
from utilities.set_openai_api_key import set_openai_api_key
from utilities.update_things import update_foo, update_selected_tools, update_use_embeddings, \
update_embeddings
# Temporarily address Wolfram Alpha SSL certificate issue
ssl._create_default_https_context = ssl._create_unverified_context
chat = ChatWrapper()
with gr.Blocks(css=".gradio-container {background-color: lightgray}") as block:
llm_state = gr.State()
history_state = gr.State()
chain_state = gr.State()
express_chain_state = gr.State()
tools_list_state = gr.State(TOOLS_DEFAULT_LIST)
trace_chain_state = gr.State(False)
speak_text_state = gr.State(False)
talking_head_state = gr.State(True)
monologue_state = gr.State(False) # Takes the input and repeats it back to themp4 user, optionally transforming it.
force_translate_state = gr.State(FORCE_TRANSLATE_DEFAULT) #
memory_state = gr.State()
# Pertains to Express-inator functionality
num_words_state = gr.State(NUM_WORDS_DEFAULT)
formality_state = gr.State(FORMALITY_DEFAULT)
anticipation_level_state = gr.State(EMOTION_DEFAULT)
joy_level_state = gr.State(EMOTION_DEFAULT)
trust_level_state = gr.State(EMOTION_DEFAULT)
fear_level_state = gr.State(EMOTION_DEFAULT)
surprise_level_state = gr.State(EMOTION_DEFAULT)
sadness_level_state = gr.State(EMOTION_DEFAULT)
disgust_level_state = gr.State(EMOTION_DEFAULT)
anger_level_state = gr.State(EMOTION_DEFAULT)
lang_level_state = gr.State(LANG_LEVEL_DEFAULT)
translate_to_state = gr.State(TRANSLATE_TO_DEFAULT)
literary_style_state = gr.State(LITERARY_STYLE_DEFAULT)
# Pertains to WHISPER functionality
whisper_lang_state = gr.State(WHISPER_DETECT_LANG)
# Pertains to question answering functionality
embeddings_state = gr.State()
qa_chain_state = gr.State()
docsearch_state = gr.State()
use_embeddings_state = gr.State(False)
use_gpt4_state = gr.State(USE_GPT4_DEFAULT)
with gr.Tab("Chat"):
with gr.Row():
with gr.Column():
gr.HTML(
"""
GPT + WolframAlpha + Whisper
Hit Enter after pasting your OpenAI API key.
If you have GPT-4 access, optionally select it in Settings tab.
""")
load_dotenv()
openai_api_key_textbox = gr.Textbox(placeholder="Paste your OpenAI API key (sk-...) and hit Enter",
value=os.environ.get("OPENAI_API_KEY", ""),
show_label=False,
lines=1, type='password')
with gr.Row():
with gr.Column(scale=1, min_width=TALKING_HEAD_WIDTH, visible=True):
speak_text_cb = gr.Checkbox(label="Enable speech", value=False)
speak_text_cb.change(update_foo, inputs=[speak_text_cb, speak_text_state],
outputs=[speak_text_state])
my_file = gr.File(label="Upload a file", type="file", visible=False)
# tmp_file_url = "/file=" + tmp_file.value['name']
html_video = create_html_video(LOOPING_TALKING_HEAD_VIDEO_PATH, TALKING_HEAD_WIDTH)
video_html = gr.HTML(html_video)
# my_aud_file = gr.File(label="Audio file", type="file", visible=True)
tmp_aud_file = gr.File("assets/audios/tempfile.mp3", visible=False)
tmp_aud_file_url = "/file=" + tmp_aud_file.value['name']
htm_audio = f''
audio_html = gr.HTML(htm_audio)
with gr.Column(scale=7):
chatbot = gr.Chatbot()
with gr.Row():
message = gr.Textbox(label="What's on your mind??",
placeholder="What's the answer to life, the universe, and everything?",
lines=1)
submit = gr.Button(value="Send", variant="secondary").style(full_width=False)
# UNCOMMENT TO USE WHISPER
with gr.Row():
audio_comp = gr.Microphone(source="microphone", type="filepath", label="Just say it!",
interactive=True, streaming=False)
audio_comp.change(transcribe, inputs=[audio_comp, whisper_lang_state], outputs=[message])
# TEMPORARY FOR TESTING
# with gr.Row():
# audio_comp_tb = gr.Textbox(label="Just say it!", lines=1)
# audio_comp_tb.submit(transcribe_dummy, inputs=[audio_comp_tb, whisper_lang_state], outputs=[message])
gr.Examples(
examples=["How many people live in Canada?",
"What is 2 to the 30th power?",
"If x+y=10 and x-y=4, what are x and y?",
"How much did it rain in SF today?",
"Get me information about the movie 'Avatar'",
"What are the top tech headlines in the US?",
"On the desk, you see two blue booklets, two purple booklets, and two yellow pairs of sunglasses - "
"if I remove all the pairs of sunglasses from the desk, how many purple items remain on it?"],
inputs=message
)
with gr.Tab("Settings"):
tools_checkbox_group = gr.CheckboxGroup(label="Tools:", choices=TOOLS_LIST,
value=TOOLS_DEFAULT_LIST)
tools_checkbox_group.change(update_selected_tools,
inputs=[tools_checkbox_group, tools_list_state, llm_state],
outputs=[tools_list_state, llm_state, chain_state, express_chain_state])
trace_chain_checkbox = gr.Checkbox(label="Show reasoning chain in chat bubble", value=False)
trace_chain_checkbox.change(update_foo, inputs=[trace_chain_checkbox, trace_chain_state],
outputs=[trace_chain_state])
force_translate_checkbox = gr.Checkbox(label="Force translation to selected Output Language",
value=FORCE_TRANSLATE_DEFAULT)
force_translate_checkbox.change(update_foo, inputs=[force_translate_checkbox, force_translate_state],
outputs=[force_translate_state])
# speak_text_cb = gr.Checkbox(label="Speak text from agent", value=False)
# speak_text_cb.change(update_foo, inputs=[speak_text_cb, speak_text_state],
# outputs=[speak_text_state])
talking_head_cb = gr.Checkbox(label="Show talking head", value=True)
talking_head_cb.change(update_talking_head, inputs=[talking_head_cb, talking_head_state],
outputs=[talking_head_state, video_html])
monologue_cb = gr.Checkbox(label="Babel fish mode (translate/restate what you enter, no conversational agent)",
value=False)
monologue_cb.change(update_foo, inputs=[monologue_cb, monologue_state],
outputs=[monologue_state])
use_gpt4_cb = gr.Checkbox(label="Use GPT-4 (experimental) if your OpenAI API has access to it",
value=USE_GPT4_DEFAULT)
use_gpt4_cb.change(set_openai_api_key,
inputs=[openai_api_key_textbox, use_gpt4_cb],
outputs=[chain_state, express_chain_state, llm_state, embeddings_state,
qa_chain_state, memory_state, use_gpt4_state])
reset_btn = gr.Button(value="Reset chat", variant="secondary").style(full_width=False)
reset_btn.click(reset_memory, inputs=[history_state, memory_state],
outputs=[chatbot, history_state, memory_state])
with gr.Tab("Whisper STT"):
whisper_lang_radio = gr.Radio(label="Whisper speech-to-text language:", choices=[
WHISPER_DETECT_LANG, "Arabic", "Arabic (Gulf)", "Catalan", "Chinese (Cantonese)", "Chinese (Mandarin)",
"Danish", "Dutch", "English (Australian)", "English (British)", "English (Indian)", "English (New Zealand)",
"English (South African)", "English (US)", "English (Welsh)", "Finnish", "French", "French (Canadian)",
"German", "German (Austrian)", "Georgian", "Hindi", "Icelandic", "Indonesian", "Italian", "Japanese",
"Korean", "Norwegian", "Polish",
"Portuguese (Brazilian)", "Portuguese (European)", "Romanian", "Russian", "Spanish (European)",
"Spanish (Mexican)", "Spanish (US)", "Swedish", "Turkish", "Ukrainian", "Welsh"],
value=WHISPER_DETECT_LANG)
whisper_lang_radio.change(update_foo,
inputs=[whisper_lang_radio, whisper_lang_state],
outputs=[whisper_lang_state])
with gr.Tab("Output Language"):
lang_level_radio = gr.Radio(label="Language level:", choices=[
LANG_LEVEL_DEFAULT, "1st grade", "2nd grade", "3rd grade", "4th grade", "5th grade", "6th grade",
"7th grade", "8th grade", "9th grade", "10th grade", "11th grade", "12th grade", "University"],
value=LANG_LEVEL_DEFAULT)
lang_level_radio.change(update_foo, inputs=[lang_level_radio, lang_level_state],
outputs=[lang_level_state])
translate_to_radio = gr.Radio(label="Language:", choices=[
TRANSLATE_TO_DEFAULT, "Arabic", "Arabic (Gulf)", "Catalan", "Chinese (Cantonese)", "Chinese (Mandarin)",
"Danish", "Dutch", "English (Australian)", "English (British)", "English (Indian)", "English (New Zealand)",
"English (South African)", "English (US)", "English (Welsh)", "Finnish", "French", "French (Canadian)",
"German", "German (Austrian)", "Georgian", "Hindi", "Icelandic", "Indonesian", "Italian", "Japanese",
"Korean", "Norwegian", "Polish",
"Portuguese (Brazilian)", "Portuguese (European)", "Romanian", "Russian", "Spanish (European)",
"Spanish (Mexican)", "Spanish (US)", "Swedish", "Turkish", "Ukrainian", "Welsh",
"emojis", "Gen Z slang", "how the stereotypical Karen would say it", "Klingon", "Neanderthal",
"Pirate", "Strange Planet expospeak technical talk", "Yoda"],
value=TRANSLATE_TO_DEFAULT)
translate_to_radio.change(update_foo,
inputs=[translate_to_radio, translate_to_state],
outputs=[translate_to_state])
with gr.Tab("Formality"):
formality_radio = gr.Radio(label="Formality:",
choices=[FORMALITY_DEFAULT, "Casual", "Polite", "Honorific"],
value=FORMALITY_DEFAULT)
formality_radio.change(update_foo,
inputs=[formality_radio, formality_state],
outputs=[formality_state])
with gr.Tab("Lit Style"):
literary_style_radio = gr.Radio(label="Literary style:", choices=[
LITERARY_STYLE_DEFAULT, "Prose", "Story", "Summary", "Outline", "Bullets", "Poetry", "Haiku", "Limerick",
"Rap",
"Joke", "Knock-knock", "FAQ"],
value=LITERARY_STYLE_DEFAULT)
literary_style_radio.change(update_foo,
inputs=[literary_style_radio, literary_style_state],
outputs=[literary_style_state])
with gr.Tab("Emotions"):
anticipation_level_radio = gr.Radio(label="Anticipation level:",
choices=[EMOTION_DEFAULT, "Interest", "Anticipation", "Vigilance"],
value=EMOTION_DEFAULT)
anticipation_level_radio.change(update_foo,
inputs=[anticipation_level_radio, anticipation_level_state],
outputs=[anticipation_level_state])
joy_level_radio = gr.Radio(label="Joy level:",
choices=[EMOTION_DEFAULT, "Serenity", "Joy", "Ecstasy"],
value=EMOTION_DEFAULT)
joy_level_radio.change(update_foo,
inputs=[joy_level_radio, joy_level_state],
outputs=[joy_level_state])
trust_level_radio = gr.Radio(label="Trust level:",
choices=[EMOTION_DEFAULT, "Acceptance", "Trust", "Admiration"],
value=EMOTION_DEFAULT)
trust_level_radio.change(update_foo,
inputs=[trust_level_radio, trust_level_state],
outputs=[trust_level_state])
fear_level_radio = gr.Radio(label="Fear level:",
choices=[EMOTION_DEFAULT, "Apprehension", "Fear", "Terror"],
value=EMOTION_DEFAULT)
fear_level_radio.change(update_foo,
inputs=[fear_level_radio, fear_level_state],
outputs=[fear_level_state])
surprise_level_radio = gr.Radio(label="Surprise level:",
choices=[EMOTION_DEFAULT, "Distraction", "Surprise", "Amazement"],
value=EMOTION_DEFAULT)
surprise_level_radio.change(update_foo,
inputs=[surprise_level_radio, surprise_level_state],
outputs=[surprise_level_state])
sadness_level_radio = gr.Radio(label="Sadness level:",
choices=[EMOTION_DEFAULT, "Pensiveness", "Sadness", "Grief"],
value=EMOTION_DEFAULT)
sadness_level_radio.change(update_foo,
inputs=[sadness_level_radio, sadness_level_state],
outputs=[sadness_level_state])
disgust_level_radio = gr.Radio(label="Disgust level:",
choices=[EMOTION_DEFAULT, "Boredom", "Disgust", "Loathing"],
value=EMOTION_DEFAULT)
disgust_level_radio.change(update_foo,
inputs=[disgust_level_radio, disgust_level_state],
outputs=[disgust_level_state])
anger_level_radio = gr.Radio(label="Anger level:",
choices=[EMOTION_DEFAULT, "Annoyance", "Anger", "Rage"],
value=EMOTION_DEFAULT)
anger_level_radio.change(update_foo,
inputs=[anger_level_radio, anger_level_state],
outputs=[anger_level_state])
with gr.Tab("Max Words"):
num_words_slider = gr.Slider(label="Max number of words to generate (0 for don't care)",
value=NUM_WORDS_DEFAULT, minimum=0, maximum=MAX_WORDS, step=10)
num_words_slider.change(update_foo,
inputs=[num_words_slider, num_words_state],
outputs=[num_words_state])
with gr.Tab("Embeddings"):
embeddings_text_box = gr.Textbox(label="Enter text for embeddings and hit Create:",
lines=20)
with gr.Row():
use_embeddings_cb = gr.Checkbox(label="Use embeddings", value=False)
use_embeddings_cb.change(update_use_embeddings, inputs=[use_embeddings_cb, use_embeddings_state],
outputs=[use_embeddings_state])
embeddings_text_submit = gr.Button(value="Create", variant="secondary").style(full_width=False)
embeddings_text_submit.click(update_embeddings,
inputs=[embeddings_text_box, embeddings_state, qa_chain_state],
outputs=[docsearch_state])
gr.HTML("""
This application, developed by James L. Weaver,
demonstrates a conversational agent implemented with OpenAI GPT-3.5 and LangChain.
When necessary, it leverages tools for complex math, searching the internet, and accessing news and weather.
Uses talking heads from Ex-Human.
For faster inference without waiting in queue, you may duplicate the space.