app.py

import together
import os

# set your API key
together.api_key = "c9909567768fbf1a69fbd94c758e432f0a05a6755c32dced992ac6640a8cfd79"

# list available models and descriptons
models = together.Models.list()

together.Models.start("togethercomputer/llama-2-7b-chat")

from langchain.llms import Together


llm = Together(
    model="togethercomputer/llama-2-7b-chat",
    temperature=0.7,
    max_tokens=128,
    top_k=1,
    together_api_key="c9909567768fbf1a69fbd94c758e432f0a05a6755c32dced992ac6640a8cfd79"
)

from langchain.vectorstores import Chroma
from langchain.text_splitter import RecursiveCharacterTextSplitter

from langchain.chains import RetrievalQA
from langchain.document_loaders import TextLoader
from langchain.document_loaders import PyPDFLoader
from langchain.document_loaders import DirectoryLoader

pdf_path = os.path.join(os.path.dirname(__file__), 'Production-Table - Sheet1 (2).pdf')

loader = PyPDFLoader(pdf_path)

documents = loader.load()
#splitting the text into
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
texts = text_splitter.split_documents(documents)
from langchain.embeddings import HuggingFaceBgeEmbeddings

model_name = "BAAI/bge-base-en"
encode_kwargs = {'normalize_embeddings': True} # set True to compute cosine similarity

model_norm = HuggingFaceBgeEmbeddings(
    model_name=model_name,
    model_kwargs={'device': 'cuda'},
    encode_kwargs=encode_kwargs
)


# Embed and store the texts
# Supplying a persist_directory will store the embeddings on disk

persist_directory = 'db'

## Here is the nmew embeddings being used
embedding = model_norm

vectordb = Chroma.from_documents(documents=texts,
                                 embedding=embedding,
                                 persist_directory=persist_directory)

retriever = vectordb.as_retriever(search_kwargs={"k": 5})
## Default LLaMA-2 prompt style
B_INST, E_INST = "[INST]", "[/INST]"
B_SYS, E_SYS = "<<SYS>>\n", "\n<</SYS>>\n\n"
DEFAULT_SYSTEM_PROMPT = """\
You are a helpful, respectful and honest assistant of a production company. You should honestly answer the user's query using the knowledge of the company's production documents uploaded.

If a question does not make any sense, or is not factually coherent, explain why instead of answering something not correct. If you don't know the answer to a question, please don't share false information."""

def get_prompt(instruction, new_system_prompt=DEFAULT_SYSTEM_PROMPT ):
    SYSTEM_PROMPT = B_SYS + new_system_prompt + E_SYS
    prompt_template =  B_INST + SYSTEM_PROMPT + instruction + E_INST
    return prompt_template

sys_prompt = """You are a helpful, respectful and honest assistant of a production company. You should honestly answer the user's query using the knowledge of the company's production documents uploaded.

If a question does not make any sense, or is not factually coherent, explain why instead of answering something not correct. If you don't know the answer to a question, please don't share false information."""

instruction = """CONTEXT:/n/n {context}/n

Question: {question}"""
get_prompt(instruction, sys_prompt)

from langchain.prompts import PromptTemplate
prompt_template = get_prompt(instruction, sys_prompt)

llama_prompt = PromptTemplate(
    template=prompt_template, input_variables=["context", "question"]
)
chain_type_kwargs = {"prompt": llama_prompt}
from langchain.schema import prompt
# create the chain to answer questions
qa_chain = RetrievalQA.from_chain_type(llm=llm,
                                       chain_type="stuff",
                                       retriever=retriever,
                                       chain_type_kwargs=chain_type_kwargs,
                                       return_source_documents=True)
## Cite sources

import textwrap

def wrap_text_preserve_newlines(text, width=110):
    # Split the input text into lines based on newline characters
    lines = text.split('\n')

    # Wrap each line individually
    wrapped_lines = [textwrap.fill(line, width=width) for line in lines]

    # Join the wrapped lines back together using newline characters
    wrapped_text = '\n'.join(wrapped_lines)

    return wrapped_text

def process_llm_response(llm_response):
    print(wrap_text_preserve_newlines(llm_response['result']))
    print('\n\nSources:')
    for source in llm_response["source_documents"]:
        print(source.metadata['source'])
import gradio as gr

with gr.Blocks() as demo:
    chatbot = gr.Chatbot()
    msg = gr.Textbox()
    clear = gr.Button("Clear")

    def user(user_message, history):
        return "", history + [[user_message, None]]

    def bot(history):
        print("Question: ", history[-1][0])
        #wrap_text_preserve_newlines(llm_response['result'])
        #bot_message = process_llm_response(qa_chain(history[-1][0]))
        bot_message = wrap_text_preserve_newlines((qa_chain(history[-1][0]))['result'])
        print("Response: ", bot_message)
        history[-1][1] = ""
        history[-1][1] += bot_message
        return history


    msg.submit(user, [msg, chatbot], [msg, chatbot], queue=False).then(bot, chatbot, chatbot)
    clear.click(lambda: None, None, chatbot, queue=False)

demo.queue()
demo.launch(debug = True)