|
1
|
!pip install -qU langchain_experimental langchain_community langchain fastembed pypdf langchain-groq chromadb
|
|
2
|
%pip install --upgrade --quiet langchain langchain-community
|
|
3
|
%pip install langchain_openai
|
|
4
|
|
|
5
|
from langchain.document_loaders import PyPDFLoader
|
|
6
|
from langchain.text_splitter import RecursiveCharacterTextSplitter
|
|
7
|
loader = PyPDFLoader("hesc103.pdf")
|
|
8
|
documents = loader.load()
|
|
9
|
|
|
10
|
from langchain.text_splitter import RecursiveCharacterTextSplitter
|
|
11
|
|
|
12
|
text_splitter = RecursiveCharacterTextSplitter(
|
|
13
|
chunk_size=1000,
|
|
14
|
chunk_overlap=100,
|
|
15
|
length_function=len,
|
|
16
|
is_separator_regex=False
|
|
17
|
)
|
|
18
|
naive_chunks = text_splitter.split_documents(documents)
|
|
19
|
|
|
20
|
|
|
21
|
from langchain_community.embeddings.fastembed import FastEmbedEmbeddings
|
|
22
|
embed_model = FastEmbedEmbeddings(model_name="BAAI/bge-base-en-v1.5")
|
|
23
|
|
|
24
|
from google.colab import userdata
|
|
25
|
from langchain_groq import ChatGroq
|
|
26
|
groq_api_key = "key"
|
|
27
|
|
|
28
|
from langchain_community.vectorstores import Chroma
|
|
29
|
naive_chunk_vectorstore = Chroma.from_documents(naive_chunks, embedding=embed_model)
|
|
30
|
|
|
31
|
from langchain_core.prompts import ChatPromptTemplate
|
|
32
|
|
|
33
|
rag_template = """\
|
|
34
|
Answer the question based only on the following Context:
|
|
35
|
{context}
|
|
36
|
|
|
37
|
Question:
|
|
38
|
{question}
|
|
39
|
"""
|
|
40
|
|
|
41
|
rag_prompt = ChatPromptTemplate.from_template(rag_template)
|
|
42
|
|
|
43
|
from langchain.chains import ConversationChain
|
|
44
|
# from langchain.llms import OpenAI
|
|
45
|
from langchain.memory import ConversationBufferMemory
|
|
46
|
|
|
47
|
chat_model = ChatGroq(temperature=0,
|
|
48
|
model_name="mixtral-8x7b-32768",
|
|
49
|
api_key="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",)
|
|
50
|
|
|
51
|
from langchain_core.runnables import RunnablePassthrough
|
|
52
|
from langchain_core.output_parsers import StrOutputParser
|
|
53
|
|
|
54
|
naive_rag_chain = (
|
|
55
|
{"context" : naive_chunk_retriever, "question" : RunnablePassthrough()}
|
|
56
|
| rag_prompt
|
|
57
|
| chat_model
|
|
58
|
| StrOutputParser()
|
|
59
|
)
|
|
60
|
memory = ConversationBufferMemory()
|
|
61
|
conversation = ConversationChain(llm=naive_rag_chain, memory=memory, verbose=True)
|
|
62
|
|
|
63
|
while True:
|
|
64
|
user_input = input("You: ")
|
|
65
|
|
|
66
|
if user_input.lower() == "exit":
|
|
67
|
break
|
|
68
|
|
|
69
|
response = conversation.predict(input=user_input)
|
|
70
|
print("AI: ", response)
|
|
71
|
|
|
72
|
|