Spaces:
Runtime error
Runtime error
Carlos Salgado
commited on
Commit
•
dbd084e
1
Parent(s):
2050d2a
update prompt template, minor fixes
Browse files- DocVerifyRAG.py +0 -148
- backend/generate_metadata.py +3 -3
- ingest.py +0 -7
DocVerifyRAG.py
DELETED
@@ -1,148 +0,0 @@
|
|
1 |
-
import streamlit as st
|
2 |
-
from dotenv import load_dotenv
|
3 |
-
from PyPDF2 import PdfReader
|
4 |
-
from langchain import embeddings
|
5 |
-
from langchain.text_splitter import CharacterTextSplitter
|
6 |
-
from langchain.embeddings import OpenAIEmbeddings, HuggingFaceInstructEmbeddings
|
7 |
-
from langchain.vectorstores import FAISS
|
8 |
-
from langchain.vectorstores import faiss
|
9 |
-
from langchain.chat_models import ChatOpenAI
|
10 |
-
from langchain.memory import ConversationBufferMemory
|
11 |
-
from langchain.chains import ConversationalRetrievalChain
|
12 |
-
from html_templates import css, bot_template, user_template
|
13 |
-
from langchain.llms import HuggingFaceHub
|
14 |
-
import os
|
15 |
-
import pickle
|
16 |
-
from datetime import datetime
|
17 |
-
|
18 |
-
|
19 |
-
def get_pdf_text(pdf_docs):
|
20 |
-
text = ""
|
21 |
-
for pdf in pdf_docs:
|
22 |
-
pdf_reader = PdfReader(pdf)
|
23 |
-
for page in pdf_reader.pages:
|
24 |
-
text += page.extract_text()
|
25 |
-
return text
|
26 |
-
|
27 |
-
|
28 |
-
def get_text_chunks(text):
|
29 |
-
text_splitter = CharacterTextSplitter(
|
30 |
-
separator="\n",
|
31 |
-
chunk_size=1000,
|
32 |
-
chunk_overlap=200,
|
33 |
-
length_function=len
|
34 |
-
)
|
35 |
-
chunks = text_splitter.split_text(text)
|
36 |
-
return chunks
|
37 |
-
|
38 |
-
|
39 |
-
def get_vectorstore(text_chunks):
|
40 |
-
embeddings = OpenAIEmbeddings()
|
41 |
-
# embeddings = HuggingFaceInstructEmbeddings(model_name="hkunlp/instructor-xl")
|
42 |
-
vectorstore = FAISS.from_texts(texts=text_chunks, embedding=embeddings)
|
43 |
-
return vectorstore
|
44 |
-
|
45 |
-
|
46 |
-
def get_conversation_chain(vectorstore):
|
47 |
-
llm = ChatOpenAI()
|
48 |
-
# llm = HuggingFaceHub(repo_id="google/flan-t5-xxl", model_kwargs={"temperature":0.5, "max_length":512})
|
49 |
-
|
50 |
-
memory = ConversationBufferMemory(
|
51 |
-
memory_key='chat_history', return_messages=True)
|
52 |
-
conversation_chain = ConversationalRetrievalChain.from_llm(
|
53 |
-
llm=llm,
|
54 |
-
retriever=vectorstore.as_retriever(),
|
55 |
-
memory=memory
|
56 |
-
)
|
57 |
-
return conversation_chain
|
58 |
-
|
59 |
-
|
60 |
-
def handle_userinput(user_question):
|
61 |
-
response = st.session_state.conversation({'question': user_question})
|
62 |
-
st.session_state.chat_history = response['chat_history']
|
63 |
-
|
64 |
-
for i, message in enumerate(st.session_state.chat_history):
|
65 |
-
# Display user message
|
66 |
-
if i % 2 == 0:
|
67 |
-
st.write(user_template.replace("{{MSG}}", message.content), unsafe_allow_html=True)
|
68 |
-
else:
|
69 |
-
print(message)
|
70 |
-
# Display AI response
|
71 |
-
st.write(bot_template.replace("{{MSG}}", message.content), unsafe_allow_html=True)
|
72 |
-
# Display source document information if available in the message
|
73 |
-
if hasattr(message, 'source') and message.source:
|
74 |
-
st.write(f"Source Document: {message.source}", unsafe_allow_html=True)
|
75 |
-
|
76 |
-
|
77 |
-
def safe_vec_store():
|
78 |
-
os.makedirs('vectorstore', exist_ok=True)
|
79 |
-
filename = 'vectores' + datetime.now().strftime('%Y%m%d%H%M') + '.pkl'
|
80 |
-
file_path = os.path.join('vectorstore', filename)
|
81 |
-
vector_store = st.session_state.vectorstore
|
82 |
-
|
83 |
-
# Serialize and save the entire FAISS object using pickle
|
84 |
-
with open(file_path, 'wb') as f:
|
85 |
-
pickle.dump(vector_store, f)
|
86 |
-
|
87 |
-
|
88 |
-
def main():
|
89 |
-
load_dotenv()
|
90 |
-
st.set_page_config(page_title="DOC Verify RAG", page_icon=":hospital:")
|
91 |
-
st.write(css, unsafe_allow_html=True)
|
92 |
-
|
93 |
-
st.subheader("Your documents")
|
94 |
-
pdf_docs = st.file_uploader("Upload your PDFs here and click on 'Process'", accept_multiple_files=True)
|
95 |
-
filenames = [file.name for file in pdf_docs if file is not None]
|
96 |
-
|
97 |
-
if "conversation" not in st.session_state:
|
98 |
-
st.session_state.conversation = None
|
99 |
-
if "chat_history" not in st.session_state:
|
100 |
-
st.session_state.chat_history = None
|
101 |
-
|
102 |
-
st.header("DOC Verify RAG :hospital:")
|
103 |
-
user_question = st.text_input("Ask a question about your documents:")
|
104 |
-
if user_question:
|
105 |
-
handle_userinput(user_question)
|
106 |
-
|
107 |
-
with st.sidebar:
|
108 |
-
|
109 |
-
st.subheader("Classification Instrucitons")
|
110 |
-
classifier_docs = st.file_uploader("Upload your instructions here and click on 'Process'", accept_multiple_files=True)
|
111 |
-
filenames = [file.name for file in classifier_docs if file is not None]
|
112 |
-
|
113 |
-
if st.button("Process"):
|
114 |
-
with st.spinner("Processing"):
|
115 |
-
loaded_vec_store = None
|
116 |
-
for filename in filenames:
|
117 |
-
if ".pkl" in filename:
|
118 |
-
file_path = os.path.join('vectorstore', filename)
|
119 |
-
with open(file_path, 'rb') as f:
|
120 |
-
loaded_vec_store = pickle.load(f)
|
121 |
-
raw_text = get_pdf_text(pdf_docs)
|
122 |
-
text_chunks = get_text_chunks(raw_text)
|
123 |
-
vec = get_vectorstore(text_chunks)
|
124 |
-
if loaded_vec_store:
|
125 |
-
vec.merge_from(loaded_vec_store)
|
126 |
-
st.warning("loaded vectorstore")
|
127 |
-
if "vectorstore" in st.session_state:
|
128 |
-
vec.merge_from(st.session_state.vectorstore)
|
129 |
-
st.warning("merged to existing")
|
130 |
-
st.session_state.vectorstore = vec
|
131 |
-
st.session_state.conversation = get_conversation_chain(vec)
|
132 |
-
st.success("data loaded")
|
133 |
-
|
134 |
-
# Save and Load Embeddings
|
135 |
-
if st.button("Save Embeddings"):
|
136 |
-
if "vectorstore" in st.session_state:
|
137 |
-
safe_vec_store()
|
138 |
-
# st.session_state.vectorstore.save_local("faiss_index")
|
139 |
-
st.sidebar.success("safed")
|
140 |
-
else:
|
141 |
-
st.sidebar.warning("No embeddings to save. Please process documents first.")
|
142 |
-
|
143 |
-
if st.button("Load Embeddings"):
|
144 |
-
st.warning("this function is not in use, just upload the vectorstore")
|
145 |
-
|
146 |
-
|
147 |
-
if __name__ == '__main__':
|
148 |
-
main()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
backend/generate_metadata.py
CHANGED
@@ -91,9 +91,9 @@ def extract_metadata(docs):
|
|
91 |
}
|
92 |
]
|
93 |
)
|
|
|
|
|
94 |
|
95 |
-
created_user = json.loads(chat_completion.choices[0].message.content)
|
96 |
-
return created_user
|
97 |
|
98 |
if __name__ == "__main__":
|
99 |
parser = argparse.ArgumentParser(description="Generate metadata for a BIM document")
|
@@ -108,4 +108,4 @@ if __name__ == "__main__":
|
|
108 |
|
109 |
docs = ingest(args.document)
|
110 |
metadata = extract_metadata(docs)
|
111 |
-
print(
|
|
|
91 |
}
|
92 |
]
|
93 |
)
|
94 |
+
# returns a dictionary
|
95 |
+
return json.loads(chat_completion.choices[0].message.content)
|
96 |
|
|
|
|
|
97 |
|
98 |
if __name__ == "__main__":
|
99 |
parser = argparse.ArgumentParser(description="Generate metadata for a BIM document")
|
|
|
108 |
|
109 |
docs = ingest(args.document)
|
110 |
metadata = extract_metadata(docs)
|
111 |
+
print(metadata)
|
ingest.py
DELETED
@@ -1,7 +0,0 @@
|
|
1 |
-
from langchain_community.document_loaders import UnstructuredPDFLoader
|
2 |
-
|
3 |
-
def ingest_pdf(path):
|
4 |
-
loader = UnstructuredPDFLoader()
|
5 |
-
text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=0)
|
6 |
-
|
7 |
-
return data
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|