sanjoy19 commited on
Commit
6d8fd97
·
verified ·
1 Parent(s): dc85378

Create app.py

Browse files
Files changed (1) hide show
  1. app.py +66 -0
app.py ADDED
@@ -0,0 +1,66 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import gradio as gr
2
+
3
+ # LangChain imports
4
+ from langchain_community.document_loaders import PyPDFLoader
5
+ from langchain_text_splitters import RecursiveCharacterTextSplitter
6
+ from langchain_community.embeddings import HuggingFaceEmbeddings
7
+ from langchain_community.vectorstores import FAISS
8
+
9
+ # LLM
10
+ from transformers import pipeline
11
+ from langchain_community.llms import HuggingFacePipeline
12
+
13
+ # -----------------------------
14
+ # Step 1: Load docs
15
+ loader1 = PyPDFLoader("AST-1.pdf")
16
+ loader2 = PyPDFLoader("AST-2.pdf")
17
+ docs = loader1.load() + loader2.load()
18
+
19
+ # Step 2: Split
20
+ splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=100)
21
+ chunks = splitter.split_documents(docs)
22
+
23
+ # Step 3: Embeddings + DB
24
+ embedding = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
25
+ db = FAISS.from_documents(chunks, embedding)
26
+
27
+ # Step 4: Retriever (MMR)
28
+ retriever = db.as_retriever(search_type="mmr", search_kwargs={"k": 4})
29
+
30
+ # Step 5: LLM
31
+ pipe = pipeline("text-generation", model="google/flan-t5-base", max_new_tokens=200)
32
+ llm = HuggingFacePipeline(pipeline=pipe)
33
+
34
+ # RAG function
35
+ def rag_query(query):
36
+ docs = retriever.invoke(query)
37
+ context = "\n\n".join([d.page_content for d in docs])
38
+
39
+ prompt = f"""
40
+ Answer the question based only on the context below.
41
+
42
+ Context:
43
+ {context}
44
+
45
+ Question:
46
+ {query}
47
+ """
48
+
49
+ return llm.invoke(prompt)
50
+
51
+ # -----------------------------
52
+ # Gradio UI
53
+ with gr.Blocks() as app:
54
+ gr.Markdown("# 📚 RAG Q&A System")
55
+
56
+ chatbot = gr.Chatbot(height=500)
57
+ msg = gr.Textbox(placeholder="Ask your question...")
58
+
59
+ def respond(message, history):
60
+ response = rag_query(message)
61
+ history.append((message, response))
62
+ return "", history
63
+
64
+ msg.submit(respond, [msg, chatbot], [msg, chatbot])
65
+
66
+ app.launch()