feat: remove cache, add context expander
Browse files- .gitignore +2 -1
- app.py +13 -7
- llm.py +38 -13
- vector_store.py +6 -8
.gitignore
CHANGED
@@ -1,3 +1,4 @@
|
|
1 |
/__pycache__
|
2 |
/temp
|
3 |
-
/models
|
|
|
|
1 |
/__pycache__
|
2 |
/temp
|
3 |
+
/models
|
4 |
+
/chroma
|
app.py
CHANGED
@@ -1,4 +1,5 @@
|
|
1 |
import os
|
|
|
2 |
import streamlit as st
|
3 |
from llm import load_llm, response_generator
|
4 |
from vector_store import load_vector_store, process_pdf
|
@@ -9,20 +10,26 @@ from uuid import uuid4
|
|
9 |
repo_id = "Qwen/Qwen2.5-3B-Instruct-GGUF"
|
10 |
filename = "qwen2.5-3b-instruct-q5_k_m.gguf"
|
11 |
|
|
|
12 |
llm = load_llm(repo_id, filename)
|
|
|
|
|
13 |
|
14 |
st.title("PDF QA")
|
15 |
# Initialize chat history
|
16 |
if "messages" not in st.session_state:
|
|
|
|
|
|
|
17 |
st.session_state.messages = []
|
18 |
|
19 |
# Display chat messages from history on app rerun
|
20 |
for message in st.session_state.messages:
|
21 |
with st.chat_message(message["role"]):
|
22 |
if message["role"] == "user":
|
23 |
-
st.
|
24 |
else:
|
25 |
-
st.
|
26 |
|
27 |
# Accept user input
|
28 |
if prompt := st.chat_input("What is up?"):
|
@@ -34,13 +41,12 @@ if prompt := st.chat_input("What is up?"):
|
|
34 |
|
35 |
# Display assistant response in chat message container
|
36 |
with st.chat_message("assistant"):
|
37 |
-
|
38 |
-
retriever = vector_store.as_retriever()
|
39 |
-
docs = retriever.get_relevant_documents(prompt)
|
40 |
-
|
41 |
response = response_generator(llm, st.session_state.messages, prompt, retriever)
|
42 |
|
43 |
st.markdown(response["answer"])
|
|
|
|
|
44 |
|
45 |
# Add assistant response to chat history
|
46 |
st.session_state.messages.append(
|
@@ -54,7 +60,7 @@ with st.sidebar:
|
|
54 |
"Choose a PDF file", accept_multiple_files=True, type="pdf"
|
55 |
)
|
56 |
if uploaded_files is not None:
|
57 |
-
|
58 |
for uploaded_file in uploaded_files:
|
59 |
temp_dir = "./temp"
|
60 |
if not os.path.exists(temp_dir):
|
|
|
1 |
import os
|
2 |
+
import shutil
|
3 |
import streamlit as st
|
4 |
from llm import load_llm, response_generator
|
5 |
from vector_store import load_vector_store, process_pdf
|
|
|
10 |
repo_id = "Qwen/Qwen2.5-3B-Instruct-GGUF"
|
11 |
filename = "qwen2.5-3b-instruct-q5_k_m.gguf"
|
12 |
|
13 |
+
|
14 |
llm = load_llm(repo_id, filename)
|
15 |
+
vector_store = load_vector_store()
|
16 |
+
|
17 |
|
18 |
st.title("PDF QA")
|
19 |
# Initialize chat history
|
20 |
if "messages" not in st.session_state:
|
21 |
+
vector_store.reset_collection()
|
22 |
+
if os.path.exists("./temp"):
|
23 |
+
shutil.rmtree("./temp")
|
24 |
st.session_state.messages = []
|
25 |
|
26 |
# Display chat messages from history on app rerun
|
27 |
for message in st.session_state.messages:
|
28 |
with st.chat_message(message["role"]):
|
29 |
if message["role"] == "user":
|
30 |
+
st.write(message["content"])
|
31 |
else:
|
32 |
+
st.write(message["content"])
|
33 |
|
34 |
# Accept user input
|
35 |
if prompt := st.chat_input("What is up?"):
|
|
|
41 |
|
42 |
# Display assistant response in chat message container
|
43 |
with st.chat_message("assistant"):
|
44 |
+
retriever = vector_store.as_retriever(search_kwargs={"k": 3})
|
|
|
|
|
|
|
45 |
response = response_generator(llm, st.session_state.messages, prompt, retriever)
|
46 |
|
47 |
st.markdown(response["answer"])
|
48 |
+
with st.expander("See context"):
|
49 |
+
st.write(response["context"])
|
50 |
|
51 |
# Add assistant response to chat history
|
52 |
st.session_state.messages.append(
|
|
|
60 |
"Choose a PDF file", accept_multiple_files=True, type="pdf"
|
61 |
)
|
62 |
if uploaded_files is not None:
|
63 |
+
st.session_state.uploaded_pdf = True
|
64 |
for uploaded_file in uploaded_files:
|
65 |
temp_dir = "./temp"
|
66 |
if not os.path.exists(temp_dir):
|
llm.py
CHANGED
@@ -7,6 +7,10 @@ from langchain.chains import create_retrieval_chain
|
|
7 |
from langchain.chains.combine_documents import create_stuff_documents_chain
|
8 |
from langchain_core.prompts import ChatPromptTemplate
|
9 |
|
|
|
|
|
|
|
|
|
10 |
|
11 |
@st.cache_resource()
|
12 |
def load_llm(repo_id, filename):
|
@@ -29,6 +33,8 @@ def load_llm(repo_id, filename):
|
|
29 |
n_threads=4,
|
30 |
n_threads_batch=4,
|
31 |
n_ctx=8000,
|
|
|
|
|
32 |
)
|
33 |
print(f"{repo_id} loaded successfully. ✅")
|
34 |
return llm
|
@@ -36,26 +42,45 @@ def load_llm(repo_id, filename):
|
|
36 |
|
37 |
# Streamed response emulator
|
38 |
def response_generator(llm, messages, question, retriever):
|
|
|
39 |
system_prompt = (
|
40 |
-
"
|
41 |
-
"
|
42 |
-
"the
|
43 |
-
"
|
44 |
-
"
|
45 |
"\n\n"
|
46 |
-
"{context}"
|
|
|
47 |
)
|
48 |
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
55 |
|
|
|
56 |
question_answer_chain = create_stuff_documents_chain(llm, prompt)
|
57 |
rag_chain = create_retrieval_chain(retriever, question_answer_chain)
|
58 |
|
59 |
-
|
|
|
60 |
|
61 |
return results
|
|
|
7 |
from langchain.chains.combine_documents import create_stuff_documents_chain
|
8 |
from langchain_core.prompts import ChatPromptTemplate
|
9 |
|
10 |
+
from langchain_core.globals import set_debug
|
11 |
+
|
12 |
+
set_debug(True)
|
13 |
+
|
14 |
|
15 |
@st.cache_resource()
|
16 |
def load_llm(repo_id, filename):
|
|
|
33 |
n_threads=4,
|
34 |
n_threads_batch=4,
|
35 |
n_ctx=8000,
|
36 |
+
max_tokens=128,
|
37 |
+
# stop=["."],
|
38 |
)
|
39 |
print(f"{repo_id} loaded successfully. ✅")
|
40 |
return llm
|
|
|
42 |
|
43 |
# Streamed response emulator
|
44 |
def response_generator(llm, messages, question, retriever):
|
45 |
+
# System prompt setting up context for the assistant
|
46 |
system_prompt = (
|
47 |
+
"<|im_start|>system\n"
|
48 |
+
"You are an AI assistant specializing in question-answering tasks. "
|
49 |
+
"Utilize the provided context and past conversation to answer "
|
50 |
+
"the current question. If the answer is unknown, clearly state that you "
|
51 |
+
"don't know. Keep responses concise and direct."
|
52 |
"\n\n"
|
53 |
+
"Context: {context}"
|
54 |
+
"\n<|im_end|>"
|
55 |
)
|
56 |
|
57 |
+
# Prepare message history
|
58 |
+
message_history = [("system", system_prompt)]
|
59 |
+
|
60 |
+
# Append conversation history to messages
|
61 |
+
for message in messages:
|
62 |
+
if message["role"] == "user":
|
63 |
+
message_history.append(
|
64 |
+
("user", "<|im_start|>user\n" + message["content"] + "\n<|im_end|>")
|
65 |
+
)
|
66 |
+
elif message["role"] == "assistant":
|
67 |
+
message_history.append(
|
68 |
+
(
|
69 |
+
"assistant",
|
70 |
+
"<|im_start|>assistant\n" + message["content"] + "\n<|im_end|>",
|
71 |
+
)
|
72 |
+
)
|
73 |
+
|
74 |
+
message_history.append(("assistant", "<|im_start|>assistant\n"))
|
75 |
+
|
76 |
+
# Create prompt template with full message history
|
77 |
+
prompt = ChatPromptTemplate.from_messages(message_history)
|
78 |
|
79 |
+
# Instantiate chains for document retrieval and question answering
|
80 |
question_answer_chain = create_stuff_documents_chain(llm, prompt)
|
81 |
rag_chain = create_retrieval_chain(retriever, question_answer_chain)
|
82 |
|
83 |
+
# Invoke RAG (retrieval-augmented generation) chain with current input
|
84 |
+
results = rag_chain.invoke({"input": question}, verbose=True)
|
85 |
|
86 |
return results
|
vector_store.py
CHANGED
@@ -1,7 +1,8 @@
|
|
1 |
import streamlit as st
|
2 |
-
|
3 |
from langchain_huggingface import HuggingFaceEmbeddings
|
4 |
from langchain_chroma import Chroma
|
|
|
5 |
from langchain_community.document_loaders import PyPDFLoader
|
6 |
from langchain_text_splitters import RecursiveCharacterTextSplitter
|
7 |
|
@@ -16,17 +17,14 @@ def load_embedding_model(model):
|
|
16 |
return model
|
17 |
|
18 |
|
|
|
19 |
def load_vector_store():
|
20 |
-
""
|
21 |
-
|
22 |
-
I didn't use @st.cache because I want to
|
23 |
-
load vector store on every page load
|
24 |
-
"""
|
25 |
-
model = load_embedding_model("sentence-transformers/all-MiniLM-L6-v2")
|
26 |
-
chromadb.api.client.SharedSystemClient.clear_system_cache()
|
27 |
vector_store = Chroma(
|
28 |
collection_name="main_store",
|
29 |
embedding_function=model,
|
|
|
30 |
)
|
31 |
return vector_store
|
32 |
|
|
|
1 |
import streamlit as st
|
2 |
+
|
3 |
from langchain_huggingface import HuggingFaceEmbeddings
|
4 |
from langchain_chroma import Chroma
|
5 |
+
from langchain_community.vectorstores import InMemoryVectorStore
|
6 |
from langchain_community.document_loaders import PyPDFLoader
|
7 |
from langchain_text_splitters import RecursiveCharacterTextSplitter
|
8 |
|
|
|
17 |
return model
|
18 |
|
19 |
|
20 |
+
@st.cache_resource()
|
21 |
def load_vector_store():
|
22 |
+
model = load_embedding_model("sentence-transformers/all-mpnet-base-v2")
|
23 |
+
|
|
|
|
|
|
|
|
|
|
|
24 |
vector_store = Chroma(
|
25 |
collection_name="main_store",
|
26 |
embedding_function=model,
|
27 |
+
persist_directory="./chroma",
|
28 |
)
|
29 |
return vector_store
|
30 |
|