Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -8,8 +8,8 @@ from langchain_core.runnables import RunnablePassthrough
|
|
8 |
from langchain_text_splitters import RecursiveCharacterTextSplitter
|
9 |
import bs4
|
10 |
import torch
|
11 |
-
import
|
12 |
-
|
13 |
# APP Title
|
14 |
st.title("Blog Retrieval and Question Answering")
|
15 |
|
@@ -21,12 +21,11 @@ api_key_Groq = st.text_input("Enter your Groq_API_KEY", type="password")
|
|
21 |
|
22 |
# Check if both API keys have been provided
|
23 |
if not api_key_langchain or not api_key_Groq:
|
24 |
-
st.write("Please enter both API keys
|
25 |
else:
|
26 |
st.write("Both API keys are set.")
|
27 |
|
28 |
# Initialize the LLM with the provided Groq API key
|
29 |
-
from langchain_groq import ChatGroq
|
30 |
llm = ChatGroq(model="llama3-8b-8192", groq_api_key=api_key_Groq)
|
31 |
|
32 |
# Define the embedding class
|
@@ -49,45 +48,55 @@ else:
|
|
49 |
# Initialize the embedding class
|
50 |
embedding_model = SentenceTransformerEmbedding('all-MiniLM-L6-v2')
|
51 |
|
52 |
-
#
|
53 |
-
|
54 |
-
loader = WebBaseLoader(
|
55 |
-
web_paths=("https://lilianweng.github.io/posts/2023-06-23-agent/",),
|
56 |
-
bs_kwargs=dict(
|
57 |
-
parse_only=bs4.SoupStrainer(
|
58 |
-
class_=("post-content", "post-title", "post-header")
|
59 |
-
)
|
60 |
-
),
|
61 |
-
)
|
62 |
-
docs = loader.load()
|
63 |
-
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
|
64 |
-
splits = text_splitter.split_documents(docs)
|
65 |
-
vectorstore = Chroma.from_documents(documents=splits, embedding=embedding_model)
|
66 |
-
return vectorstore
|
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 |
-
|
|
|
|
|
|
8 |
from langchain_text_splitters import RecursiveCharacterTextSplitter
|
9 |
import bs4
|
10 |
import torch
|
11 |
+
from langchain_groq import ChatGroq
|
12 |
+
|
13 |
# APP Title
|
14 |
st.title("Blog Retrieval and Question Answering")
|
15 |
|
|
|
21 |
|
22 |
# Check if both API keys have been provided
|
23 |
if not api_key_langchain or not api_key_Groq:
|
24 |
+
st.write("Please enter both API keys to access this APP.")
|
25 |
else:
|
26 |
st.write("Both API keys are set.")
|
27 |
|
28 |
# Initialize the LLM with the provided Groq API key
|
|
|
29 |
llm = ChatGroq(model="llama3-8b-8192", groq_api_key=api_key_Groq)
|
30 |
|
31 |
# Define the embedding class
|
|
|
48 |
# Initialize the embedding class
|
49 |
embedding_model = SentenceTransformerEmbedding('all-MiniLM-L6-v2')
|
50 |
|
51 |
+
# Streamlit UI for blog URL input
|
52 |
+
blog_url = st.text_input("Enter the URL of the blog to retrieve:")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
53 |
|
54 |
+
# Load, chunk, and index the contents of the blog
|
55 |
+
def load_data(url):
|
56 |
+
try:
|
57 |
+
loader = WebBaseLoader(
|
58 |
+
web_paths=(url,),
|
59 |
+
bs_kwargs=dict(
|
60 |
+
parse_only=bs4.SoupStrainer(
|
61 |
+
class_=("post-content", "post-title", "post-header")
|
62 |
+
)
|
63 |
+
),
|
64 |
+
)
|
65 |
+
docs = loader.load()
|
66 |
+
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
|
67 |
+
splits = text_splitter.split_documents(docs)
|
68 |
+
vectorstore = Chroma.from_documents(documents=splits, embedding=embedding_model)
|
69 |
+
return vectorstore
|
70 |
+
except Exception as e:
|
71 |
+
st.error(f"An error occurred while loading the blog: {e}")
|
72 |
+
return None
|
73 |
|
74 |
+
# Load the data if a URL is provided
|
75 |
+
if blog_url:
|
76 |
+
vectorstore = load_data(blog_url)
|
77 |
+
if vectorstore:
|
78 |
+
# Streamlit UI for question input
|
79 |
+
question = st.text_input("Enter your question:")
|
80 |
|
81 |
+
if question:
|
82 |
+
retriever = vectorstore.as_retriever()
|
83 |
+
prompt = hub.pull("rlm/rag-prompt", api_key=api_key_langchain)
|
84 |
|
85 |
+
def format_docs(docs):
|
86 |
+
return "\n\n".join(doc.page_content for doc in docs)
|
87 |
|
88 |
+
rag_chain = (
|
89 |
+
{"context": retriever | format_docs, "question": RunnablePassthrough()}
|
90 |
+
| prompt
|
91 |
+
| llm
|
92 |
+
| StrOutputParser()
|
93 |
+
)
|
94 |
|
95 |
+
# Example invocation
|
96 |
+
try:
|
97 |
+
result = rag_chain.invoke(question)
|
98 |
+
st.write("Answer:", result)
|
99 |
+
except Exception as e:
|
100 |
+
st.error(f"An error occurred while generating the answer: {e}")
|
101 |
+
else:
|
102 |
+
st.write("Failed to load the blog content. Please check the URL and try again.")
|