Trading-Chatbot / app.py
vjain's picture
Update app.py
dbb5ddb
raw
history blame
4.12 kB
import gradio as gr
import openai
import pandas as pd
import numpy as np
import csv
import os
from datasets import load_dataset
openai.api_key= os.environ.get("openai.api_key")
from openai.embeddings_utils import get_embedding
from openai.embeddings_utils import cosine_similarity
import requests
model_id = "sentence-transformers/all-MiniLM-L6-v2"
import json
hf_token = os.environ.get("hf_token")
import re
from sklearn.metrics.pairwise import cosine_similarity
def generate_embeddings(texts, model_id, hf_token):
api_url = f"https://api-inference.huggingface.co/pipeline/feature-extraction/{model_id}"
headers = {"Authorization": f"Bearer {hf_token}"}
response = requests.post(api_url, headers=headers, json={"inputs": texts, "options":{"wait_for_model":True}})
embeddings = response.json()
return embeddings
AP_Bio = load_dataset('vjain/biology_AP_embeddings')
df1 = pd.DataFrame(AP_Bio['train'])
df1["similarity"] = 0
AP_Physics = load_dataset('vjain/AP_physics_embeddings')
df2 = pd.DataFrame(AP_Physics['train'])
df2["similarity"] = 0
dataframes = {
"AP_Bio": df1,
"AP_Physics": df2
}
#df = pd.read_csv("TA_embeddings.csv")
#df["embedding"]=df["embedding"].apply(eval).apply(np.array)
def reply(input, dataset_name):
global messages
try:
if dataset_name not in dataframes:
return "Invalid dataset selected. Please select a valid dataset."
if not input:
return "Please Enter a Question to get an Answer"
df = dataframes[dataset_name]
input = input
input_vector = generate_embeddings(input, model_id,hf_token)
df["similarities"]=df["embedding"].apply(lambda x: cosine_similarity([x],[input_vector])[0][0])
data = df.sort_values("similarities", ascending=False).head(10)
data.to_csv("sorted.csv")
context = []
for i, row in data.iterrows():
context.append(row['text'])
context
text = "\n".join(context)
context = text
prompt = f"""
Answer the following question using the context given below.If you don't know the answer for certain, say I don't know.
Context: {context}
Q: {input}
"""
response= openai.Completion.create(
prompt=prompt,
temperature=1,
max_tokens=500,
top_p=1,
frequency_penalty=0,
presence_penalty=0,
model="text-davinci-003"
)["choices"][0]["text"].strip(" \n")
chat_transcript = ""
chat_transcript += input + "\n\n" + message + "\n\n"
return chat_transcript
except Exception as e:
return f"An error occurred: {e}"
csv_dropdown = gr.inputs.Dropdown(
label="Select the Book",
choices=["AP_Bio", "AP_Physics"],
default="AP_Bio"
)
input_text = gr.inputs.Textbox(
label="Enter your questions here",
placeholder="E.g. What is DNA?",
lines=3
)
text_output = gr.outputs.Textbox(label="Answer")
description = "Scholar Bot is a question answering system designed to provide accurate and relevant answers to questions from this book hosted by OpenStax https://openstax.org/details/books/biology-ap-courses. Simply enter your question in the text box above and Scholar Bot will use advanced natural language processing algorithms to search a large corpus of biology text to find the best answer for you. Scholar Bot uses the Sentence Transformers model to generate embeddings of text, and OpenAI's GPT-3 language model to provide answers to your questions."
ui = gr.Interface(fn=reply,
inputs=[input_text, csv_dropdown],
outputs=[text_output],
title="Scholar Bot",
description=description,
theme="light",
layout="vertical",
allow_flagging=False,
examples=[["What is the function of DNA polymerase?", "AP_Bio"]]
)
ui.launch()