Spaces:
Sleeping
Sleeping
Update app.py
#14
by Muthuraja18 - opened
app.py
CHANGED
|
@@ -1,31 +1,40 @@
|
|
| 1 |
import streamlit as st
|
|
|
|
|
|
|
| 2 |
|
| 3 |
-
#
|
| 4 |
from langchain_community.document_loaders import PyPDFLoader, TextLoader
|
| 5 |
from langchain_text_splitters import RecursiveCharacterTextSplitter
|
| 6 |
from langchain_community.embeddings import HuggingFaceEmbeddings
|
| 7 |
from langchain_community.vectorstores import FAISS
|
| 8 |
from langchain_community.llms import HuggingFacePipeline
|
| 9 |
-
|
| 10 |
from langchain.chains import RetrievalQA
|
|
|
|
| 11 |
from transformers import pipeline
|
| 12 |
|
| 13 |
|
| 14 |
# -------------------------------
|
| 15 |
-
# Load Documents
|
| 16 |
# -------------------------------
|
| 17 |
def load_documents(uploaded_files):
|
| 18 |
documents = []
|
|
|
|
| 19 |
for file in uploaded_files:
|
| 20 |
-
|
| 21 |
-
f.write(file.getbuffer())
|
| 22 |
|
| 23 |
-
|
| 24 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 25 |
else:
|
| 26 |
-
loader = TextLoader(
|
| 27 |
|
| 28 |
documents.extend(loader.load())
|
|
|
|
|
|
|
|
|
|
| 29 |
return documents
|
| 30 |
|
| 31 |
|
|
@@ -51,19 +60,21 @@ def create_vectorstore(chunks):
|
|
| 51 |
|
| 52 |
|
| 53 |
# -------------------------------
|
| 54 |
-
# Load Local LLM (
|
| 55 |
# -------------------------------
|
|
|
|
| 56 |
def load_llm():
|
| 57 |
pipe = pipeline(
|
| 58 |
-
"text2text-generation",
|
| 59 |
model="google/flan-t5-base",
|
| 60 |
-
max_length=512
|
|
|
|
| 61 |
)
|
| 62 |
return HuggingFacePipeline(pipeline=pipe)
|
| 63 |
|
| 64 |
|
| 65 |
# -------------------------------
|
| 66 |
-
# Build QA Chain
|
| 67 |
# -------------------------------
|
| 68 |
def build_qa(vectorstore):
|
| 69 |
llm = load_llm()
|
|
@@ -94,12 +105,12 @@ if uploaded_files:
|
|
| 94 |
vectorstore = create_vectorstore(chunks)
|
| 95 |
qa_chain = build_qa(vectorstore)
|
| 96 |
|
| 97 |
-
st.success("Documents ready!")
|
| 98 |
|
| 99 |
query = st.text_input("Ask a question from your documents")
|
| 100 |
|
| 101 |
if query:
|
| 102 |
with st.spinner("Generating answer..."):
|
| 103 |
-
result = qa_chain.
|
| 104 |
st.write("### Answer:")
|
| 105 |
-
st.write(result)
|
|
|
|
| 1 |
import streamlit as st
|
| 2 |
+
import tempfile
|
| 3 |
+
import os
|
| 4 |
|
| 5 |
+
# LangChain imports (new structure)
|
| 6 |
from langchain_community.document_loaders import PyPDFLoader, TextLoader
|
| 7 |
from langchain_text_splitters import RecursiveCharacterTextSplitter
|
| 8 |
from langchain_community.embeddings import HuggingFaceEmbeddings
|
| 9 |
from langchain_community.vectorstores import FAISS
|
| 10 |
from langchain_community.llms import HuggingFacePipeline
|
|
|
|
| 11 |
from langchain.chains import RetrievalQA
|
| 12 |
+
|
| 13 |
from transformers import pipeline
|
| 14 |
|
| 15 |
|
| 16 |
# -------------------------------
|
| 17 |
+
# Load Documents (FIXED temp file handling)
|
| 18 |
# -------------------------------
|
| 19 |
def load_documents(uploaded_files):
|
| 20 |
documents = []
|
| 21 |
+
|
| 22 |
for file in uploaded_files:
|
| 23 |
+
suffix = file.name.split(".")[-1]
|
|
|
|
| 24 |
|
| 25 |
+
with tempfile.NamedTemporaryFile(delete=False, suffix=f".{suffix}") as tmp:
|
| 26 |
+
tmp.write(file.getbuffer())
|
| 27 |
+
tmp_path = tmp.name
|
| 28 |
+
|
| 29 |
+
if suffix == "pdf":
|
| 30 |
+
loader = PyPDFLoader(tmp_path)
|
| 31 |
else:
|
| 32 |
+
loader = TextLoader(tmp_path)
|
| 33 |
|
| 34 |
documents.extend(loader.load())
|
| 35 |
+
|
| 36 |
+
os.remove(tmp_path) # cleanup
|
| 37 |
+
|
| 38 |
return documents
|
| 39 |
|
| 40 |
|
|
|
|
| 60 |
|
| 61 |
|
| 62 |
# -------------------------------
|
| 63 |
+
# Load Local LLM (STABLE VERSION)
|
| 64 |
# -------------------------------
|
| 65 |
+
@st.cache_resource
|
| 66 |
def load_llm():
|
| 67 |
pipe = pipeline(
|
| 68 |
+
task="text2text-generation",
|
| 69 |
model="google/flan-t5-base",
|
| 70 |
+
max_length=512,
|
| 71 |
+
do_sample=False
|
| 72 |
)
|
| 73 |
return HuggingFacePipeline(pipeline=pipe)
|
| 74 |
|
| 75 |
|
| 76 |
# -------------------------------
|
| 77 |
+
# Build QA Chain
|
| 78 |
# -------------------------------
|
| 79 |
def build_qa(vectorstore):
|
| 80 |
llm = load_llm()
|
|
|
|
| 105 |
vectorstore = create_vectorstore(chunks)
|
| 106 |
qa_chain = build_qa(vectorstore)
|
| 107 |
|
| 108 |
+
st.success("✅ Documents ready!")
|
| 109 |
|
| 110 |
query = st.text_input("Ask a question from your documents")
|
| 111 |
|
| 112 |
if query:
|
| 113 |
with st.spinner("Generating answer..."):
|
| 114 |
+
result = qa_chain.invoke({"query": query})
|
| 115 |
st.write("### Answer:")
|
| 116 |
+
st.write(result["result"])
|