push for posterity

app.py

@@ -1,20 +1,14 @@
-import io
 import os
 import ssl
-from contextlib import closing
 
-import boto3
 import gradio as gr
-import requests
-# Pertains to Express-inator functionality
-from langchain.prompts import PromptTemplate
+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, TOOLS_LIST, MAX_WORDS
+    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.audio import POLLY_VOICE_DATA, AZURE_VOICE_DATA
 from utilities.html_stuff import create_html_video, update_talking_head
-from utilities.polly_utils import NEURAL_ENGINE
 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, \
@@ -23,8 +17,6 @@ from utilities.update_things import update_foo, update_selected_tools, update_us
 # 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:
@@ -36,7 +28,7 @@ with gr.Blocks(css=".gradio-container {background-color: lightgray}") as block:
     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 the user, optionally transforming it.
+    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()
 
@@ -74,8 +66,11 @@ with gr.Blocks(css=".gradio-container {background-color: lightgray}") as block:
                     <p><center>Hit Enter after pasting your OpenAI API key.</center></p>
                     <i><center>If you have GPT-4 access, optionally select it in Settings tab.</center></i>""")
 
+            load_dotenv()
             openai_api_key_textbox = gr.Textbox(placeholder="Paste your OpenAI API key (sk-...) and hit Enter",
-                                                show_label=False, lines=1, type='password')
+                                                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):
@@ -86,11 +81,11 @@ with gr.Blocks(css=".gradio-container {background-color: lightgray}") as block:
                 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, TALKING_HEAD_WIDTH)
+                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("audios/tempfile.mp3", visible=False)
+                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><source src={tmp_aud_file_url} type="audio/mp3"></audio>'
                 audio_html = gr.HTML(htm_audio)

config/config.py

@@ -1,5 +1,3 @@
-import os
-
 from langchain import PromptTemplate
 
 TALKING_HEAD_WIDTH = 192
@@ -12,12 +10,12 @@ TOOLS_LIST = ['serpapi', 'wolfram-alpha', 'pal-math',
 TOOLS_DEFAULT_LIST = ['serpapi']
 BUG_FOUND_MSG = "Congratulations, you've found a bug in this application!"
 # AUTH_ERR_MSG = "Please paste your OpenAI key from openai.com to use this application. It is not necessary to hit a button or key after pasting it."
-AUTH_ERR_MSG = "Please paste your OpenAI key from openai.com to use this application. "
+AUTHORIZATION_ERROR_MESSAGE = "Please paste your OpenAI key from openai.com to use this application. "
 MAX_TOKENS = 512
 
-LOOPING_TALKING_HEAD = "videos/Masahiro.mp4"
+LOOPING_TALKING_HEAD_VIDEO_PATH = "assets/videos/Masahiro.mp4"
 
-MAX_TALKING_HEAD_TEXT_LENGTH = 155
+MAX_TALKING_HEAD_TEXT_LENGTH = 1550
 
 # Pertains to Express-inator functionality
 NUM_WORDS_DEFAULT = 0

utilities/html_stuff.py

@@ -4,16 +4,15 @@ from contextlib import closing
 
 import boto3
 import gradio as gr
-
 import requests
 
-from config.config import TALKING_HEAD_WIDTH, LOOPING_TALKING_HEAD
+from config.config import TALKING_HEAD_WIDTH, LOOPING_TALKING_HEAD_VIDEO_PATH
 from utilities.audio import AZURE_VOICE_DATA, POLLY_VOICE_DATA
 from utilities.polly_utils import NEURAL_ENGINE
 
 
 def create_html_video(file_name, width):
-    tmp_file = gr.File(LOOPING_TALKING_HEAD, visible=False)
+    tmp_file = gr.File(LOOPING_TALKING_HEAD_VIDEO_PATH, visible=False)
     temp_file_url = "/file=" + tmp_file.value['name']
     html_video = f'<video width={width} height={width} autoplay muted loop><source src={temp_file_url} type="video/mp4" poster="Masahiro.png"></video>'
     return html_video
@@ -22,7 +21,7 @@ def update_talking_head(widget, state):
     if widget:
         state = widget
 
-        video_html_talking_head = create_html_video(LOOPING_TALKING_HEAD, TALKING_HEAD_WIDTH)
+        video_html_talking_head = create_html_video(LOOPING_TALKING_HEAD_VIDEO_PATH, TALKING_HEAD_WIDTH)
         return state, video_html_talking_head
     else:
         # return state, create_html_video(LOOPING_TALKING_HEAD, "32")
@@ -60,9 +59,9 @@ def do_html_audio_speak(words_to_speak, polly_language):
             # output = os.path.join("/tmp/", "speech.mp3")
 
             try:
-                with open('audios/tempfile.mp3', 'wb') as f:
+                with open('assets/audios/tempfile.mp3', 'wb') as f:
                     f.write(stream.read())
-                temp_aud_file = gr.File("audios/tempfile.mp3")
+                temp_aud_file = gr.File("assets/audios/tempfile.mp3")
                 temp_aud_file_url = "/file=" + temp_aud_file.value['name']
                 html_audio = f'<audio autoplay><source src={temp_aud_file_url} type="audio/mp3"></audio>'
             except IOError as error:
@@ -74,7 +73,7 @@ def do_html_audio_speak(words_to_speak, polly_language):
         print("Could not stream audio")
         return None, None
 
-    return html_audio, "audios/tempfile.mp3"
+    return html_audio, "assets/audios/tempfile.mp3"
 
 
 

utilities/reset_memory.py

@@ -4,7 +4,7 @@ from typing import Tuple, Optional
 
 from langchain import ConversationChain, LLMChain
 
-from config.config import MAX_TALKING_HEAD_TEXT_LENGTH, LOOPING_TALKING_HEAD, TALKING_HEAD_WIDTH
+from config.config import MAX_TALKING_HEAD_TEXT_LENGTH, LOOPING_TALKING_HEAD_VIDEO_PATH, TALKING_HEAD_WIDTH
 from utilities.html_stuff import do_html_video_speak, create_html_video, do_html_audio_speak
 from utilities.transform_text import transform_text
 
@@ -76,14 +76,14 @@ class ChatWrapper:
                     if len(output) <= MAX_TALKING_HEAD_TEXT_LENGTH:
                         html_video, temp_file = do_html_video_speak(output, translate_to)
                     else:
-                        temp_file = LOOPING_TALKING_HEAD
+                        temp_file = LOOPING_TALKING_HEAD_VIDEO_PATH
                         html_video = create_html_video(temp_file, TALKING_HEAD_WIDTH)
                         html_audio, temp_aud_file = do_html_audio_speak(output, translate_to)
                 else:
                     html_audio, temp_aud_file = do_html_audio_speak(output, translate_to)
             else:
                 if talking_head:
-                    temp_file = LOOPING_TALKING_HEAD
+                    temp_file = LOOPING_TALKING_HEAD_VIDEO_PATH
                     html_video = create_html_video(temp_file, TALKING_HEAD_WIDTH)
                 else:
                     # html_audio, temp_aud_file = do_html_audio_speak(output, translate_to)

utilities/run_chain.py

@@ -6,7 +6,7 @@ from io import StringIO
 from openai import InvalidRequestError
 from openai.error import RateLimitError, AuthenticationError
 
-from config.config import BUG_FOUND_MSG, AUTH_ERR_MSG
+from config.config import BUG_FOUND_MSG, AUTHORIZATION_ERROR_MESSAGE
 
 
 def run_chain(chain, inp, capture_hidden_text):
@@ -21,7 +21,7 @@ def run_chain(chain, inp, capture_hidden_text):
         try:
             output = chain.run(input=inp)
         except AuthenticationError as ae:
-            error_msg = AUTH_ERR_MSG + str(datetime.datetime.now()) + ". " + str(ae)
+            error_msg = AUTHORIZATION_ERROR_MESSAGE + str(datetime.datetime.now()) + ". " + str(ae)
             print("error_msg", error_msg)
         except RateLimitError as rle:
             error_msg = "\n\nRateLimitError: " + str(rle)
@@ -59,7 +59,7 @@ def run_chain(chain, inp, capture_hidden_text):
         try:
             output = chain.run(input=inp)
         except AuthenticationError as ae:
-            output = AUTH_ERR_MSG + str(datetime.datetime.now()) + ". " + str(ae)
+            output = AUTHORIZATION_ERROR_MESSAGE + str(datetime.datetime.now()) + ". " + str(ae)
             print("output", output)
         except RateLimitError as rle:
             output = "\n\nRateLimitError: " + str(rle)

utilities/update_things.py

@@ -1,7 +1,6 @@
 from langchain import FAISS
 from langchain.text_splitter import CharacterTextSplitter
 
-
 from utilities.load_chain import load_chain