chatbot_tibetan / app.py
test
add verbose mode
793072f
raw
history blame
4.27 kB
import os
import time
import uuid
from typing import Dict, List, Tuple
import gradio as gr
import requests
from store import store_message_pair
# Environment Variables
DEBUG = bool(os.getenv("DEBUG", False))
VERBOSE = bool(os.getenv("V", False))
OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
BING_TRANSLATE_API_KEY = os.getenv("BING_TRANSLATE_API_KEY")
# Type Definitions
ROLE_USER = "user"
ROLE_ASSISTANT = "assistant"
CHATGPT_MSG = Dict[str, str] # {"role": "user|assistant", "content": "text"}
CHATGPT_HISTROY = List[CHATGPT_MSG]
CHATBOT_MSG = Tuple[str, str] # (user_message, bot_response)
CHATBOT_HISTORY = List[CHATBOT_MSG]
# Constants
LANG_BO = "bo"
LANG_ZH = "en"
def bing_translate(text: str, from_lang: str, to_lang: str):
if DEBUG:
if from_lang != "bo":
return "ཀཀཀཀཀཀ"
return "aaaaa"
headers = {
"Ocp-Apim-Subscription-Key": BING_TRANSLATE_API_KEY,
"Content-Type": "application/json",
"Ocp-Apim-Subscription-Region": "eastus",
"X-ClientTraceId": str(uuid.uuid4()),
}
resp = requests.post(
url="https://api.cognitive.microsofttranslator.com/translate",
params={"api-version": "3.0", "from": from_lang, "to": to_lang},
json=[{"text": text}],
headers=headers,
)
result = resp.json()
if resp.status_code == 200:
return result[0]["translations"][0]["text"]
else:
raise Exception("Error in translation API: ", result)
def make_completion(history):
if DEBUG:
time.sleep(2)
return "aaaaa"
headers = {
"Content-Type": "application/json",
"Authorization": f"Bearer {OPENAI_API_KEY}",
}
resp = requests.post(
url="https://api.openai.com/v1/chat/completions",
json={"model": "gpt-3.5-turbo", "messages": history},
headers=headers,
)
if resp.status_code == 200:
return resp.json()["choices"][0]["message"]["content"]
else:
print(resp.content)
return "Sorry, I don't understand."
def user(input_bo: str, history_bo: list):
history_bo.append([input_bo, None])
return "", history_bo
def store_chat(chat_id: str, history_bo: list, history_zh):
bo_msg_pair = history_bo[-1]
store_message_pair(chat_id, bo_msg_pair, LANG_BO)
en_msg_pair = (history_zh[-1]["content"], history_zh[-2]["content"])
store_message_pair(chat_id, en_msg_pair, LANG_ZH)
def bot(history_bo: list, history_en: list, request: gr.Request):
"""Translate user input to English, send to OpenAI, translate response to Tibetan, and return to user.
Args:
input_bo (str): Tibetan input from user
history_bo (CHATBOT_HISTORY): Tibetan history of gradio chatbot
history_en (CHATGPT_HISTORY): English history of OpenAI ChatGPT
Returns:
history_bo (CHATBOT_HISTORY): Tibetan history of gradio chatbot
history_en (CHATGPT_HISTORY): English history of OpenAI ChatGPT
"""
input_bo = history_bo[-1][0]
input_en = bing_translate(input_bo, LANG_BO, LANG_ZH)
history_en.append({"role": ROLE_USER, "content": input_en})
response_en = make_completion(history_en)
resopnse_bo = bing_translate(response_en, LANG_ZH, LANG_BO)
history_en.append({"role": ROLE_ASSISTANT, "content": response_en})
history_bo[-1][1] = resopnse_bo
if VERBOSE:
print("------------------------")
print(history_bo)
print(history_en)
print("------------------------")
store_chat(
chat_id=request.client.host, history_bo=history_bo, history_zh=history_en
)
return history_bo, history_en
with gr.Blocks() as demo:
history_en = gr.State(value=[])
history_bo = gr.Chatbot(label="Tibetan Chatbot").style(height=650)
input_bo = gr.Textbox(
show_label=False, placeholder="Type a message here and press enter"
)
input_bo.submit(
fn=user,
inputs=[input_bo, history_bo],
outputs=[input_bo, history_bo],
queue=False,
).then(
fn=bot,
inputs=[history_bo, history_en],
outputs=[history_bo, history_en],
)
clear = gr.Button("New Chat")
clear.click(lambda: [[], []], None, [history_bo, history_en], queue=False)
demo.launch()