File size: 4,089 Bytes
93461a2
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e25d199
93461a2
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
cc6c7e9
93461a2
 
 
 
 
396d74c
93461a2
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
aba63c5
93461a2
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
506fac0
f753eb9
 
 
 
 
 
6d78ad1
93461a2
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
import os
from langchain_community.embeddings.sentence_transformer import SentenceTransformerEmbeddings
from langchain_community.vectorstores import Chroma
from langchain.chains import RetrievalQA
from langchain.prompts import PromptTemplate
from langchain.text_splitter import RecursiveCharacterTextSplitter
import streamlit as st
from together import Together
from langchain.llms.base import LLM
from typing import Any, List, Optional
from Components.para_utility import load_pdfs_from_file
from pydantic import PrivateAttr

embedding_function = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2")


class TogetherLLM(LLM):
    model_name: str = "mistralai/Mistral-7B-Instruct-v0.2"
    temperature: float = 0
    max_tokens: int = 256
    together_api_key: str = os.getenv("Together_API")

    # Define private attribute
    client: Any = PrivateAttr()
    
    def __init__(self, **kwargs):
        super().__init__(**kwargs)
        # Together("api_key")=self.together_api_key
        self.client = Together(api_key=self.together_api_key)

    def _call(self, prompt: str, **kwargs: Any) -> str:
        response = self.client.chat.completions.create(
            model=self.model_name,
            messages=[{"role": "user", "content": prompt}],
            temperature=self.temperature,
            max_tokens=self.max_tokens,
        )
        return response.choices[0].message.content

    @property
    def _llm_type(self) -> str:
        return "together_llm"


def split_docs(documents, chunk_size=500, chunk_overlap=10):
    text_splitter = RecursiveCharacterTextSplitter(chunk_size=chunk_size, chunk_overlap=chunk_overlap)
    docs = text_splitter.split_documents(documents)
    return docs

def initialize_model(documents):
    with st.spinner("Processing documents may take a few seconds"):
        new_pages = split_docs(documents)
        
        if not new_pages:
            st.error("No documents to process.")
            return None
        # db = Chroma.from_documents(new_pages, embedding_function)
        db = Chroma.from_documents(
            new_pages,
            embedding_function,
            persist_directory="../dataset"
        )
        db.persist()
        # Use the new TogetherLLM class
        llm = TogetherLLM()

        retriever = db.as_retriever(similarity_score_threshold=0.95, search_kwargs={"k": 5})

        prompt_template = """
        CONTEXT: {context}
        QUESTION: {question}"""

        PROMPT = PromptTemplate(template=f"[INST] {prompt_template} [/INST]", input_variables=["context", "question"])

        chain = RetrievalQA.from_chain_type(
            llm=llm,
            chain_type='stuff',
            retriever=retriever,
            input_key='query',
            return_source_documents=True,
            chain_type_kwargs={"prompt": PROMPT},
            verbose=True
        )
    st.success("Document Processed successfully!")
    return chain

class ConversationalAgent:
    def __init__(self, chain):
        self.chain = chain
        self.history = []

    def ask(self, query):
        context = " ".join([item['response'] for item in self.history])
        prompt_template = """
        CONTEXT: {context}
        QUESTION: {question}"""
        prompt = f"[INST] CONTEXT: {context} QUESTION: {query} {prompt_template} [/INST]"
        response = self.chain(query)
        result = response['result']
        # st.write(result)
        self.history.append({'query': query, 'response': result})
        return result, response['source_documents']

def process_file(uploaded_file):
    """Process the uploaded file and return an agent"""

        
    documents = load_pdfs_from_file(uploaded_file)
    if documents is None:
        return None
    chain = initialize_model(documents)
    
    return ConversationalAgent(chain)

def demo_file_load():
    dir_pre=os.getcwd()
    pre=os.path.join(dir_pre,"dataset","LLM.pdf")
    
    documents = load_pdfs_from_file(pre)
    if documents is None:
        return None
    chain = initialize_model(documents)
    
    return ConversationalAgent(chain)