Upload 3 files
Browse files- src/app_updated.py +202 -0
- src/rag_methods.py +174 -0
- src/requirements.txt +0 -0
src/app_updated.py
ADDED
@@ -0,0 +1,202 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
import os
|
3 |
+
import dotenv
|
4 |
+
import uuid
|
5 |
+
|
6 |
+
# Patch sqlite3 for Streamlit Cloud compatibility
|
7 |
+
if os.name == 'posix':
|
8 |
+
__import__('pysqlite3')
|
9 |
+
import sys
|
10 |
+
sys.modules['sqlite3'] = sys.modules.pop('pysqlite3')
|
11 |
+
|
12 |
+
from langchain.schema import HumanMessage, AIMessage
|
13 |
+
from langchain_groq import ChatGroq
|
14 |
+
|
15 |
+
from rag_methods import (
|
16 |
+
load_doc_to_db,
|
17 |
+
load_url_to_db,
|
18 |
+
stream_llm_response,
|
19 |
+
stream_llm_rag_response,
|
20 |
+
)
|
21 |
+
|
22 |
+
dotenv.load_dotenv()
|
23 |
+
|
24 |
+
# --- Custom CSS Styling ---
|
25 |
+
def apply_custom_css():
|
26 |
+
st.markdown("""
|
27 |
+
<style>
|
28 |
+
.main .block-container {
|
29 |
+
padding-top: 2rem;
|
30 |
+
padding-bottom: 2rem;
|
31 |
+
}
|
32 |
+
h1, h2, h3, h4 {
|
33 |
+
font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif;
|
34 |
+
font-weight: 600;
|
35 |
+
}
|
36 |
+
.app-title {
|
37 |
+
text-align: center;
|
38 |
+
color: #4361ee;
|
39 |
+
font-size: 2.2rem;
|
40 |
+
font-weight: 700;
|
41 |
+
margin-bottom: 1.5rem;
|
42 |
+
padding: 1rem;
|
43 |
+
border-radius: 10px;
|
44 |
+
background: linear-gradient(90deg, rgba(67, 97, 238, 0.1), rgba(58, 12, 163, 0.1));
|
45 |
+
text-shadow: 0px 0px 2px rgba(0,0,0,0.1);
|
46 |
+
}
|
47 |
+
.chat-container {
|
48 |
+
border-radius: 10px;
|
49 |
+
padding: 10px;
|
50 |
+
margin-bottom: 1rem;
|
51 |
+
}
|
52 |
+
.message-container {
|
53 |
+
padding: 0.8rem;
|
54 |
+
margin-bottom: 0.8rem;
|
55 |
+
border-radius: 8px;
|
56 |
+
}
|
57 |
+
.user-message {
|
58 |
+
background-color: rgba(67, 97, 238, 0.15);
|
59 |
+
border-left: 4px solid #4361ee;
|
60 |
+
}
|
61 |
+
.assistant-message {
|
62 |
+
background-color: rgba(58, 12, 163, 0.1);
|
63 |
+
border-left: 4px solid #3a0ca3;
|
64 |
+
}
|
65 |
+
.document-list {
|
66 |
+
background-color: rgba(67, 97, 238, 0.05);
|
67 |
+
border-radius: 8px;
|
68 |
+
padding: 0.7rem;
|
69 |
+
}
|
70 |
+
.upload-container {
|
71 |
+
border: 2px dashed rgba(67, 97, 238, 0.5);
|
72 |
+
border-radius: 10px;
|
73 |
+
padding: 1rem;
|
74 |
+
margin-bottom: 1rem;
|
75 |
+
text-align: center;
|
76 |
+
}
|
77 |
+
.status-indicator {
|
78 |
+
font-size: 0.85rem;
|
79 |
+
font-weight: 600;
|
80 |
+
padding: 0.3rem 0.7rem;
|
81 |
+
border-radius: 20px;
|
82 |
+
display: inline-block;
|
83 |
+
margin-bottom: 0.5rem;
|
84 |
+
}
|
85 |
+
.status-active {
|
86 |
+
background-color: rgba(46, 196, 182, 0.2);
|
87 |
+
color: #2EC4B6;
|
88 |
+
}
|
89 |
+
.status-inactive {
|
90 |
+
background-color: rgba(231, 111, 81, 0.2);
|
91 |
+
color: #E76F51;
|
92 |
+
}
|
93 |
+
@media screen and (max-width: 768px) {
|
94 |
+
.app-title {
|
95 |
+
font-size: 1.8rem;
|
96 |
+
padding: 0.7rem;
|
97 |
+
}
|
98 |
+
}
|
99 |
+
</style>
|
100 |
+
""", unsafe_allow_html=True)
|
101 |
+
|
102 |
+
# --- Page Setup ---
|
103 |
+
st.set_page_config(
|
104 |
+
page_title="RAG-Xpert: An Enhanced RAG Framework",
|
105 |
+
page_icon="📚",
|
106 |
+
layout="centered",
|
107 |
+
initial_sidebar_state="expanded"
|
108 |
+
)
|
109 |
+
|
110 |
+
apply_custom_css()
|
111 |
+
|
112 |
+
st.markdown('<h1 class="app-title">📚 RAG-Xpert: An Enhanced Retrieval-Augmented Generation Framework 🤖</h1>', unsafe_allow_html=True)
|
113 |
+
|
114 |
+
# --- Session Initialization ---
|
115 |
+
if "session_id" not in st.session_state:
|
116 |
+
st.session_state.session_id = str(uuid.uuid4())
|
117 |
+
if "rag_sources" not in st.session_state:
|
118 |
+
st.session_state.rag_sources = []
|
119 |
+
if "messages" not in st.session_state:
|
120 |
+
st.session_state.messages = [
|
121 |
+
{"role": "user", "content": "Hello"},
|
122 |
+
{"role": "assistant", "content": "Hi there! How can I assist you today?"}
|
123 |
+
]
|
124 |
+
|
125 |
+
# --- Sidebar ---
|
126 |
+
with st.sidebar:
|
127 |
+
st.markdown("""
|
128 |
+
<div style="
|
129 |
+
text-align: center;
|
130 |
+
padding: 1rem 0;
|
131 |
+
margin-bottom: 1.5rem;
|
132 |
+
background: linear-gradient(to right, #4361ee22, #3a0ca322);
|
133 |
+
border-radius: 10px;">
|
134 |
+
<div style="font-size: 0.85rem; color: #888;">Developed By</div>
|
135 |
+
<div style="font-size: 1.2rem; font-weight: 700; color: #4361ee;">Uditanshu Pandey</div>
|
136 |
+
</div>
|
137 |
+
""", unsafe_allow_html=True)
|
138 |
+
|
139 |
+
is_vector_db_loaded = "vector_db" in st.session_state and st.session_state.vector_db is not None
|
140 |
+
rag_status = st.toggle("Enable Knowledge Enhancement (RAG)", value=is_vector_db_loaded, key="use_rag", disabled=not is_vector_db_loaded)
|
141 |
+
|
142 |
+
if rag_status:
|
143 |
+
st.markdown('<div class="status-indicator status-active">RAG Mode: Active ✓</div>', unsafe_allow_html=True)
|
144 |
+
else:
|
145 |
+
st.markdown('<div class="status-indicator status-inactive">RAG Mode: Inactive ✗</div>', unsafe_allow_html=True)
|
146 |
+
|
147 |
+
st.toggle("Show Retrieved Context", key="debug_mode", value=False)
|
148 |
+
st.button("🧹 Clear Chat History", on_click=lambda: st.session_state.messages.clear(), type="primary")
|
149 |
+
|
150 |
+
st.markdown("<h3 style='text-align: center; color: #4361ee; margin-top: 1.5rem;'>📚 Knowledge Sources</h3>", unsafe_allow_html=True)
|
151 |
+
st.markdown('<div class="upload-container">', unsafe_allow_html=True)
|
152 |
+
st.file_uploader("📄 Upload Documents", type=["pdf", "txt", "docx", "md"], accept_multiple_files=True, on_change=load_doc_to_db, key="rag_docs")
|
153 |
+
st.markdown('</div>', unsafe_allow_html=True)
|
154 |
+
|
155 |
+
st.text_input("🌐 Add Webpage URL", placeholder="https://example.com", on_change=load_url_to_db, key="rag_url")
|
156 |
+
|
157 |
+
doc_count = len(st.session_state.rag_sources) if is_vector_db_loaded else 0
|
158 |
+
with st.expander(f"📑 Knowledge Base ({doc_count} sources)"):
|
159 |
+
if doc_count:
|
160 |
+
st.markdown('<div class="document-list">', unsafe_allow_html=True)
|
161 |
+
for i, source in enumerate(st.session_state.rag_sources):
|
162 |
+
st.markdown(f"**{i+1}.** {source}")
|
163 |
+
st.markdown('</div>', unsafe_allow_html=True)
|
164 |
+
else:
|
165 |
+
st.info("No documents added yet. Upload files or add URLs to enhance the assistant's knowledge.")
|
166 |
+
|
167 |
+
# --- Initialize LLM ---
|
168 |
+
llm_stream = ChatGroq(
|
169 |
+
model_name="meta-llama/llama-4-scout-17b-16e-instruct",
|
170 |
+
api_key=os.getenv("GROQ_API_KEY"),
|
171 |
+
temperature=0.4,
|
172 |
+
max_tokens=1024,
|
173 |
+
)
|
174 |
+
|
175 |
+
# --- Chat Display ---
|
176 |
+
st.markdown('<div class="chat-container">', unsafe_allow_html=True)
|
177 |
+
for message in st.session_state.messages:
|
178 |
+
avatar = "👤" if message["role"] == "user" else "🤖"
|
179 |
+
css_class = "user-message" if message["role"] == "user" else "assistant-message"
|
180 |
+
with st.chat_message(message["role"], avatar=avatar):
|
181 |
+
st.markdown(f'<div class="message-container {css_class}">{message["content"]}</div>', unsafe_allow_html=True)
|
182 |
+
st.markdown('</div>', unsafe_allow_html=True)
|
183 |
+
|
184 |
+
# --- User Input Handling ---
|
185 |
+
if prompt := st.chat_input("Ask me anything..."):
|
186 |
+
st.session_state.messages.append({"role": "user", "content": prompt})
|
187 |
+
with st.chat_message("user", avatar="👤"):
|
188 |
+
st.markdown(f'<div class="message-container user-message">{prompt}</div>', unsafe_allow_html=True)
|
189 |
+
|
190 |
+
with st.chat_message("assistant", avatar="🤖"):
|
191 |
+
thinking_placeholder = st.empty()
|
192 |
+
thinking_placeholder.info("Thinking... Please wait a moment.")
|
193 |
+
messages = [
|
194 |
+
HumanMessage(content=m["content"]) if m["role"] == "user" else AIMessage(content=m["content"])
|
195 |
+
for m in st.session_state.messages
|
196 |
+
]
|
197 |
+
if not st.session_state.use_rag:
|
198 |
+
thinking_placeholder.empty()
|
199 |
+
st.write_stream(stream_llm_response(llm_stream, messages))
|
200 |
+
else:
|
201 |
+
thinking_placeholder.info("Searching knowledge base... Please wait.")
|
202 |
+
st.write_stream(stream_llm_rag_response(llm_stream, messages))
|
src/rag_methods.py
ADDED
@@ -0,0 +1,174 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import dotenv
|
3 |
+
from time import time
|
4 |
+
import streamlit as st
|
5 |
+
|
6 |
+
from langchain_community.document_loaders.text import TextLoader
|
7 |
+
from langchain_community.document_loaders import (
|
8 |
+
WebBaseLoader,
|
9 |
+
PyPDFLoader,
|
10 |
+
Docx2txtLoader,
|
11 |
+
)
|
12 |
+
from langchain_community.vectorstores import Chroma
|
13 |
+
from langchain.text_splitter import RecursiveCharacterTextSplitter
|
14 |
+
from langchain_huggingface import HuggingFaceEmbeddings
|
15 |
+
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
|
16 |
+
from langchain.chains import create_history_aware_retriever, create_retrieval_chain
|
17 |
+
from langchain.chains.combine_documents import create_stuff_documents_chain
|
18 |
+
|
19 |
+
dotenv.load_dotenv()
|
20 |
+
|
21 |
+
os.environ["USER_AGENT"] = "myagent"
|
22 |
+
DB_DOCS_LIMIT = 10
|
23 |
+
|
24 |
+
# Stream non-RAG LLM response
|
25 |
+
def stream_llm_response(llm_stream, messages):
|
26 |
+
response_message = ""
|
27 |
+
for chunk in llm_stream.stream(messages):
|
28 |
+
response_message += chunk.content
|
29 |
+
yield chunk
|
30 |
+
st.session_state.messages.append({"role": "assistant", "content": response_message})
|
31 |
+
|
32 |
+
# --- Document Loading and Indexing ---
|
33 |
+
def load_doc_to_db():
|
34 |
+
if "rag_docs" in st.session_state and st.session_state.rag_docs:
|
35 |
+
docs = []
|
36 |
+
for doc_file in st.session_state.rag_docs:
|
37 |
+
if doc_file.name not in st.session_state.rag_sources:
|
38 |
+
if len(st.session_state.rag_sources) < DB_DOCS_LIMIT:
|
39 |
+
os.makedirs("source_files", exist_ok=True)
|
40 |
+
file_path = f"./source_files/{doc_file.name}"
|
41 |
+
with open(file_path, "wb") as file:
|
42 |
+
file.write(doc_file.read())
|
43 |
+
try:
|
44 |
+
if doc_file.type == "application/pdf":
|
45 |
+
loader = PyPDFLoader(file_path)
|
46 |
+
elif doc_file.name.endswith(".docx"):
|
47 |
+
loader = Docx2txtLoader(file_path)
|
48 |
+
elif doc_file.type in ["text/plain", "text/markdown"]:
|
49 |
+
loader = TextLoader(file_path)
|
50 |
+
else:
|
51 |
+
st.warning(f"Unsupported document type: {doc_file.type}")
|
52 |
+
continue
|
53 |
+
docs.extend(loader.load())
|
54 |
+
st.session_state.rag_sources.append(doc_file.name)
|
55 |
+
except Exception as e:
|
56 |
+
st.toast(f"Error loading document {doc_file.name}: {e}", icon="⚠️")
|
57 |
+
finally:
|
58 |
+
os.remove(file_path)
|
59 |
+
else:
|
60 |
+
st.error(f"Max documents reached ({DB_DOCS_LIMIT}).")
|
61 |
+
if docs:
|
62 |
+
_split_and_load_docs(docs)
|
63 |
+
st.toast(f"Documents loaded successfully.", icon="✅")
|
64 |
+
|
65 |
+
def load_url_to_db():
|
66 |
+
if "rag_url" in st.session_state and st.session_state.rag_url:
|
67 |
+
url = st.session_state.rag_url
|
68 |
+
docs = []
|
69 |
+
if url not in st.session_state.rag_sources:
|
70 |
+
if len(st.session_state.rag_sources) < DB_DOCS_LIMIT:
|
71 |
+
try:
|
72 |
+
loader = WebBaseLoader(url)
|
73 |
+
docs.extend(loader.load())
|
74 |
+
st.session_state.rag_sources.append(url)
|
75 |
+
except Exception as e:
|
76 |
+
st.error(f"Error loading from URL {url}: {e}")
|
77 |
+
if docs:
|
78 |
+
_split_and_load_docs(docs)
|
79 |
+
st.toast(f"Loaded content from URL: {url}", icon="✅")
|
80 |
+
else:
|
81 |
+
st.error(f"Max documents reached ({DB_DOCS_LIMIT}).")
|
82 |
+
|
83 |
+
def initialize_vector_db(docs):
|
84 |
+
# Initialize HuggingFace embeddings
|
85 |
+
embedding = HuggingFaceEmbeddings(
|
86 |
+
model_name="BAAI/bge-large-en-v1.5",
|
87 |
+
model_kwargs={'device': 'cpu'},
|
88 |
+
encode_kwargs={'normalize_embeddings': False}
|
89 |
+
)
|
90 |
+
|
91 |
+
# Shared persistent directory for long-term storage
|
92 |
+
persist_dir = "./chroma_persistent_db"
|
93 |
+
collection_name = "persistent_collection"
|
94 |
+
|
95 |
+
# Create the persistent Chroma vector store
|
96 |
+
vector_db = Chroma.from_documents(
|
97 |
+
documents=docs,
|
98 |
+
embedding=embedding,
|
99 |
+
persist_directory=persist_dir,
|
100 |
+
collection_name=collection_name
|
101 |
+
)
|
102 |
+
|
103 |
+
# Persist to disk
|
104 |
+
vector_db.persist()
|
105 |
+
|
106 |
+
return vector_db
|
107 |
+
|
108 |
+
|
109 |
+
def _split_and_load_docs(docs):
|
110 |
+
text_splitter = RecursiveCharacterTextSplitter(
|
111 |
+
chunk_size=1000,
|
112 |
+
chunk_overlap=200,
|
113 |
+
)
|
114 |
+
|
115 |
+
chunks = text_splitter.split_documents(docs)
|
116 |
+
|
117 |
+
if "vector_db" not in st.session_state:
|
118 |
+
st.session_state.vector_db = initialize_vector_db(chunks)
|
119 |
+
else:
|
120 |
+
st.session_state.vector_db.add_documents(chunks)
|
121 |
+
st.session_state.vector_db.persist() # Save changes
|
122 |
+
|
123 |
+
# --- RAG Chain ---
|
124 |
+
|
125 |
+
def _get_context_retriever_chain(vector_db, llm):
|
126 |
+
retriever = vector_db.as_retriever()
|
127 |
+
prompt = ChatPromptTemplate.from_messages([
|
128 |
+
MessagesPlaceholder(variable_name="messages"),
|
129 |
+
("user", "{input}"),
|
130 |
+
("user", "Given the above conversation, generate a search query to find relevant information.")
|
131 |
+
])
|
132 |
+
return create_history_aware_retriever(llm, retriever, prompt)
|
133 |
+
|
134 |
+
def get_conversational_rag_chain(llm):
|
135 |
+
retriever_chain = _get_context_retriever_chain(st.session_state.vector_db, llm)
|
136 |
+
prompt = ChatPromptTemplate.from_messages([
|
137 |
+
("system",
|
138 |
+
"""You are a helpful assistant answering the user's queries using the provided context if available.\n
|
139 |
+
{context}"""),
|
140 |
+
MessagesPlaceholder(variable_name="messages"),
|
141 |
+
("user", "{input}")
|
142 |
+
])
|
143 |
+
stuff_documents_chain = create_stuff_documents_chain(llm, prompt)
|
144 |
+
return create_retrieval_chain(retriever_chain, stuff_documents_chain)
|
145 |
+
|
146 |
+
# Stream RAG LLM response
|
147 |
+
def stream_llm_rag_response(llm_stream, messages):
|
148 |
+
rag_chain = get_conversational_rag_chain(llm_stream)
|
149 |
+
|
150 |
+
# Extract latest user input and prior messages
|
151 |
+
input_text = messages[-1].content
|
152 |
+
history = messages[:-1]
|
153 |
+
|
154 |
+
# --- DEBUG: Show context retrieved ---
|
155 |
+
if st.session_state.get("debug_mode"):
|
156 |
+
retriever = st.session_state.vector_db.as_retriever()
|
157 |
+
retrieved_docs = retriever.get_relevant_documents(input_text)
|
158 |
+
st.markdown("### 🔍 Retrieved Context (Debug Mode)")
|
159 |
+
for i, doc in enumerate(retrieved_docs):
|
160 |
+
st.markdown(f"**Chunk {i+1}:**\n```\n{doc.page_content.strip()}\n```")
|
161 |
+
|
162 |
+
response_message = "*(RAG Response)*\n"
|
163 |
+
response = rag_chain.stream({
|
164 |
+
"messages": history,
|
165 |
+
"input": input_text
|
166 |
+
})
|
167 |
+
|
168 |
+
for chunk in response:
|
169 |
+
if 'answer' in chunk:
|
170 |
+
response_message += chunk['answer']
|
171 |
+
yield chunk['answer']
|
172 |
+
|
173 |
+
st.session_state.messages.append({"role": "assistant", "content": response_message})
|
174 |
+
|
src/requirements.txt
ADDED
Binary file (6.74 kB). View file
|
|