Spaces:
Running
Running
File size: 4,004 Bytes
48ebcb5 52d8866 48ebcb5 32c932b 52d8866 32c932b |
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 |
__import__('pysqlite3')
import sys
sys.modules['sqlite3'] = sys.modules.pop('pysqlite3')
from langchain_chroma import Chroma
from langchain_openai import OpenAIEmbeddings
from langchain_core.messages import HumanMessage
from langchain.chains import create_history_aware_retriever, create_retrieval_chain
from langchain.chains.combine_documents import create_stuff_documents_chain
from langchain_chroma import Chroma
from langchain_openai import OpenAIEmbeddings
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder, PromptTemplate
from langchain_openai import ChatOpenAI
from dotenv import load_dotenv
import gradio as gr
load_dotenv()
# OpenAI's GPT model to use
GPT_VERSION = "gpt-3.5-turbo-0125"
# Prompt used to contextualize a question inside the context of a conversation
CONTEXTUALIZATION_PROMPT = """Given a chat history and the latest user question \
which might reference context in the chat history, formulate a standalone question \
which can be understood without the chat history. Do NOT answer the question, \
just reformulate it if needed and otherwise return it as is."""
# Prompt used for building an answer to a question using the retrieved context
SYSTEM_PROMPT = """You are a marketing assistant for question-answering tasks for a company called Tryolabs. \
You must answer the question using pieces of retrieved context consisting of blogposts written by Tryolabs. \
Answer the question in a friendly manner, trying to refer to Tryolabs' content to engage the user and encourage \
to make contact with the company. \
When you can, include properly formatted hyperlinks to the content. TRY providing references to different content \
within the same response. Extract the hyperlink from the context. \
If you don't know the answer to the question, encourage the user to contact us for more information. \
Tryolabs' contact email is hello@tryolabs.com. \
Keep your answers concise. \
ALWAYS respond with Markdown. \
You MUST only use information found on the pieces of context given. You MUST NOT attribute any inventions to \
Tryolabs unless explicitly mentioned on the context. \
{context}"""
llm = ChatOpenAI(model=GPT_VERSION)
vectorstore = Chroma(persist_directory="./chroma_db", embedding_function=OpenAIEmbeddings())
retriever = vectorstore.as_retriever(search_type="mmr", search_kwargs={'k': 6, 'lambda_mult': 0.3})
contextualize_q_prompt = ChatPromptTemplate.from_messages(
[
("system", CONTEXTUALIZATION_PROMPT),
MessagesPlaceholder("chat_history"),
("human", "{input}"),
]
)
history_aware_retriever = create_history_aware_retriever(llm, retriever, contextualize_q_prompt)
custom_rag_prompt = ChatPromptTemplate.from_messages(
[
("system", SYSTEM_PROMPT),
MessagesPlaceholder("chat_history"),
("human", "{input}")
]
)
document_prompt = PromptTemplate(
input_variables=['page_content', 'title', 'link'],
template="Title: {title}, Link to blogpost: {link}, Content: {page_content}"
)
qa_chain = create_stuff_documents_chain(llm, custom_rag_prompt, document_prompt=document_prompt)
rag_chain = create_retrieval_chain(history_aware_retriever, qa_chain)
def predict(message, history):
chat_history = []
for m, a in history:
chat_history.append(HumanMessage(content=m))
chat_history.append(a)
ai_response = rag_chain.invoke({"input": message, "chat_history": chat_history})
return ai_response["answer"]
theme = gr.themes.Default().set(
button_primary_background_fill_dark="#26C8A1",
button_primary_background_fill_hover_dark="#3cd6b2",
button_primary_border_color_dark="#26C8A1",
button_primary_text_color_dark="#0B0723",
background_fill_primary_dark="#0B0723",
)
gr.ChatInterface(
fn=predict,
examples=[
"What experience do you have working with pricing optimization?",
"How can I build my own MLOps pipeline?",
"How can I fine-tune my own LLMs?"
],
theme=theme
).launch() |