Update model.py
Browse files
model.py
CHANGED
@@ -1,128 +1,130 @@
|
|
1 |
-
from langchain.prompts import PromptTemplate
|
2 |
-
from langchain_community.embeddings import HuggingFaceEmbeddings
|
3 |
-
from langchain_community.vectorstores import FAISS
|
4 |
-
from langchain.llms import HuggingFacePipeline
|
5 |
-
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
|
6 |
-
from langchain.chains import RetrievalQA
|
7 |
-
|
8 |
-
|
9 |
-
import
|
10 |
-
import
|
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 |
-
model
|
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 |
-
msg.content
|
96 |
-
await msg.
|
97 |
-
|
98 |
-
|
99 |
-
|
100 |
-
|
101 |
-
|
102 |
-
|
103 |
-
|
104 |
-
|
105 |
-
|
106 |
-
|
107 |
-
|
108 |
-
|
109 |
-
|
110 |
-
|
111 |
-
|
112 |
-
|
113 |
-
|
114 |
-
|
115 |
-
|
116 |
-
sources
|
117 |
-
|
118 |
-
|
119 |
-
|
120 |
-
|
121 |
-
|
122 |
-
|
123 |
-
|
124 |
-
|
125 |
-
|
126 |
-
|
127 |
-
|
128 |
-
|
|
|
|
|
|
1 |
+
from langchain.prompts import PromptTemplate
|
2 |
+
from langchain_community.embeddings import HuggingFaceEmbeddings
|
3 |
+
from langchain_community.vectorstores import FAISS
|
4 |
+
from langchain.llms import HuggingFacePipeline
|
5 |
+
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
|
6 |
+
from langchain.chains import RetrievalQA
|
7 |
+
from download_assets import download_assets
|
8 |
+
download_assets()
|
9 |
+
import chainlit as cl
|
10 |
+
from dotenv import load_dotenv
|
11 |
+
import torch
|
12 |
+
import os
|
13 |
+
|
14 |
+
load_dotenv()
|
15 |
+
|
16 |
+
DB_FAISS_PATH = 'vectorstore/db_faiss'
|
17 |
+
|
18 |
+
# Prompt Template
|
19 |
+
custom_prompt_template = """Use the following pieces of information to answer the user's question.
|
20 |
+
If you don't know the answer, just say that you don't know, don't try to make up an answer.
|
21 |
+
|
22 |
+
Context: {context}
|
23 |
+
Question: {question}
|
24 |
+
|
25 |
+
Only return the helpful answer below and nothing else.
|
26 |
+
Helpful answer:
|
27 |
+
"""
|
28 |
+
|
29 |
+
def set_custom_prompt():
|
30 |
+
prompt = PromptTemplate(template=custom_prompt_template,
|
31 |
+
input_variables=['context', 'question'])
|
32 |
+
return prompt
|
33 |
+
|
34 |
+
# Create RetrievalQA chain
|
35 |
+
def retrieval_qa_chain(llm, prompt, db):
|
36 |
+
qa_chain = RetrievalQA.from_chain_type(
|
37 |
+
llm=llm,
|
38 |
+
chain_type='stuff',
|
39 |
+
retriever=db.as_retriever(search_kwargs={'k': 2}),
|
40 |
+
return_source_documents=True,
|
41 |
+
chain_type_kwargs={'prompt': prompt}
|
42 |
+
)
|
43 |
+
return qa_chain
|
44 |
+
|
45 |
+
# Load Hugging Face LLM
|
46 |
+
def load_llm():
|
47 |
+
# Load model and tokenizer
|
48 |
+
tokenizer = AutoTokenizer.from_pretrained("google/flan-t5-base")
|
49 |
+
model = AutoModelForSeq2SeqLM.from_pretrained(
|
50 |
+
"google/flan-t5-base",
|
51 |
+
device_map="cpu",
|
52 |
+
torch_dtype=torch.float32
|
53 |
+
)
|
54 |
+
|
55 |
+
# Create text-generation pipeline without invalid parameters
|
56 |
+
pipe = pipeline(
|
57 |
+
"text2text-generation",
|
58 |
+
model=model,
|
59 |
+
tokenizer=tokenizer,
|
60 |
+
max_new_tokens=512,
|
61 |
+
repetition_penalty=1.15
|
62 |
+
)
|
63 |
+
|
64 |
+
# Create LangChain wrapper for the pipeline
|
65 |
+
llm = HuggingFacePipeline(pipeline=pipe)
|
66 |
+
return llm
|
67 |
+
|
68 |
+
# Build full chatbot pipeline
|
69 |
+
def qa_bot():
|
70 |
+
embeddings = HuggingFaceEmbeddings(
|
71 |
+
model_name="sentence-transformers/all-MiniLM-L6-v2",
|
72 |
+
model_kwargs={'device': 'cpu'}
|
73 |
+
)
|
74 |
+
db = FAISS.load_local(
|
75 |
+
DB_FAISS_PATH,
|
76 |
+
embeddings,
|
77 |
+
allow_dangerous_deserialization=True
|
78 |
+
)
|
79 |
+
|
80 |
+
llm = load_llm()
|
81 |
+
qa_prompt = set_custom_prompt()
|
82 |
+
qa = retrieval_qa_chain(llm, qa_prompt, db)
|
83 |
+
return qa
|
84 |
+
|
85 |
+
# Run for one query (used internally)
|
86 |
+
def final_result(query):
|
87 |
+
qa_result = qa_bot()
|
88 |
+
response = qa_result({'query': query})
|
89 |
+
return response
|
90 |
+
|
91 |
+
# Chainlit UI - Start
|
92 |
+
@cl.on_chat_start
|
93 |
+
async def start():
|
94 |
+
chain = qa_bot()
|
95 |
+
msg = cl.Message(content="Starting the bot...")
|
96 |
+
await msg.send()
|
97 |
+
msg.content = "Hi, Welcome to MindMate. What is your query?"
|
98 |
+
await msg.update()
|
99 |
+
cl.user_session.set("chain", chain)
|
100 |
+
|
101 |
+
# Chainlit UI - Handle messages
|
102 |
+
@cl.on_message
|
103 |
+
async def main(message: cl.Message):
|
104 |
+
chain = cl.user_session.get("chain")
|
105 |
+
cb = cl.AsyncLangchainCallbackHandler(
|
106 |
+
stream_final_answer=True, answer_prefix_tokens=["FINAL", "ANSWER"]
|
107 |
+
)
|
108 |
+
cb.answer_reached = True
|
109 |
+
|
110 |
+
# Use invoke with proper query format
|
111 |
+
res = await cl.make_async(chain.invoke)(
|
112 |
+
{"query": message.content},
|
113 |
+
callbacks=[cb]
|
114 |
+
)
|
115 |
+
|
116 |
+
# Extract result and sources from the response
|
117 |
+
answer = res.get("result", "No result found")
|
118 |
+
sources = res.get("source_documents", [])
|
119 |
+
|
120 |
+
# Format sources to show only the content
|
121 |
+
if sources:
|
122 |
+
formatted_sources = []
|
123 |
+
for source in sources:
|
124 |
+
if hasattr(source, 'page_content'):
|
125 |
+
formatted_sources.append(source.page_content.strip())
|
126 |
+
|
127 |
+
if formatted_sources:
|
128 |
+
answer = f"{answer}\n\nBased on the following information:\n" + "\n\n".join(formatted_sources)
|
129 |
+
|
130 |
+
await cl.Message(content=answer).send()
|