Update app.py
Browse files
app.py
CHANGED
@@ -1,104 +1,85 @@
|
|
1 |
import os
|
2 |
import streamlit as st
|
3 |
-
import pandas as pd
|
4 |
-
import numpy as np
|
5 |
import faiss
|
|
|
6 |
from datasets import load_dataset
|
7 |
from sentence_transformers import SentenceTransformer
|
8 |
from groq import Groq
|
9 |
|
10 |
-
# Constants
|
11 |
-
|
12 |
-
|
|
|
|
|
13 |
|
14 |
-
#
|
15 |
-
|
16 |
-
def load_data():
|
17 |
-
dataset = load_dataset("FreedomIntelligence/RAG-Instruct", split="train")
|
18 |
-
df = pd.DataFrame(dataset)
|
19 |
-
return df[["question", "answer"]]
|
20 |
|
21 |
-
#
|
22 |
-
@st.cache_resource
|
23 |
-
def setup_faiss(data):
|
24 |
-
model = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")
|
25 |
-
|
26 |
-
if os.path.exists(INDEX_FILE) and os.path.exists(QUESTIONS_FILE):
|
27 |
-
st.info("π Loading FAISS index from disk...")
|
28 |
-
index = faiss.read_index(INDEX_FILE)
|
29 |
-
questions = np.load(QUESTIONS_FILE, allow_pickle=True)
|
30 |
-
else:
|
31 |
-
st.info("βοΈ FAISS index not found. Building new index...")
|
32 |
-
|
33 |
-
questions = data["question"].tolist()
|
34 |
-
embeddings = []
|
35 |
-
progress_bar = st.progress(0, text="Embedding questions...")
|
36 |
-
total = len(questions)
|
37 |
-
|
38 |
-
for i, chunk in enumerate(np.array_split(questions, 10)):
|
39 |
-
emb = model.encode(chunk)
|
40 |
-
embeddings.extend(emb)
|
41 |
-
progress_bar.progress((i + 1) / 10, text=f"Embedding... {int((i + 1) * 10)}%")
|
42 |
-
|
43 |
-
embeddings = np.array(embeddings)
|
44 |
-
index = faiss.IndexFlatL2(embeddings.shape[1])
|
45 |
-
index.add(embeddings)
|
46 |
-
|
47 |
-
faiss.write_index(index, INDEX_FILE)
|
48 |
-
np.save(QUESTIONS_FILE, np.array(questions, dtype=object))
|
49 |
-
|
50 |
-
progress_bar.empty()
|
51 |
-
st.success("β
FAISS index built and saved!")
|
52 |
-
|
53 |
-
return model, index, questions
|
54 |
-
|
55 |
-
|
56 |
-
# Retrieve relevant context
|
57 |
-
def retrieve_context(query, model, index, questions, data, top_k=1):
|
58 |
-
query_vec = model.encode([query])
|
59 |
-
distances, indices = index.search(np.array(query_vec), top_k)
|
60 |
-
results = [questions[i] + "\n\n" + data.iloc[i]["answer"] for i in indices[0]]
|
61 |
-
return "\n\n".join(results)
|
62 |
-
|
63 |
-
# Call Groq LLM
|
64 |
-
def query_groq(context, query):
|
65 |
-
prompt = f"Context:\n{context}\n\nQuestion: {query}\n\nAnswer:"
|
66 |
-
client = Groq(api_key=st.secrets["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"])
|
67 |
-
response = client.chat.completions.create(
|
68 |
-
messages=[{"role": "user", "content": prompt}],
|
69 |
-
model="llama-3-70b-8192"
|
70 |
-
)
|
71 |
-
return response.choices[0].message.content
|
72 |
-
|
73 |
-
# Streamlit UI
|
74 |
st.set_page_config(page_title="RAG App with Groq", layout="wide")
|
75 |
-
st.title("
|
76 |
-
|
77 |
-
# Load data and setup
|
78 |
-
data = load_data()
|
79 |
-
model, index, questions = setup_faiss(data)
|
80 |
-
|
81 |
-
st.markdown("Ask a question based on the QA knowledge base.")
|
82 |
|
83 |
-
#
|
84 |
-
|
85 |
-
|
86 |
-
"
|
87 |
-
|
88 |
-
|
89 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
90 |
]
|
91 |
|
92 |
-
|
|
|
|
|
93 |
if st.button("Ask"):
|
94 |
-
with st.spinner("Retrieving and generating
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
|
99 |
-
|
100 |
-
|
101 |
-
|
102 |
-
|
103 |
-
|
104 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
import os
|
2 |
import streamlit as st
|
|
|
|
|
3 |
import faiss
|
4 |
+
import pickle
|
5 |
from datasets import load_dataset
|
6 |
from sentence_transformers import SentenceTransformer
|
7 |
from groq import Groq
|
8 |
|
9 |
+
# Constants
|
10 |
+
DATASET_NAME = "neural-bridge/rag-dataset-1200"
|
11 |
+
MODEL_NAME = "all-MiniLM-L6-v2"
|
12 |
+
INDEX_FILE = "faiss_index.pkl"
|
13 |
+
DOCS_FILE = "contexts.pkl"
|
14 |
|
15 |
+
# Set up Groq client
|
16 |
+
client = Groq(api_key=os.environ.get("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"))
|
|
|
|
|
|
|
|
|
17 |
|
18 |
+
# UI
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
19 |
st.set_page_config(page_title="RAG App with Groq", layout="wide")
|
20 |
+
st.title("π§ Retrieval-Augmented Generation (RAG) App")
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
|
22 |
+
# Load or create vector DB
|
23 |
+
@st.cache_resource
|
24 |
+
def setup_database():
|
25 |
+
st.info("Loading dataset and setting up database...")
|
26 |
+
progress = st.progress(0)
|
27 |
+
|
28 |
+
dataset = load_dataset(DATASET_NAME, split="train")
|
29 |
+
contexts = [entry["context"] for entry in dataset]
|
30 |
+
|
31 |
+
embedder = SentenceTransformer(MODEL_NAME)
|
32 |
+
embeddings = embedder.encode(contexts, show_progress_bar=True)
|
33 |
+
|
34 |
+
dimension = embeddings[0].shape[0]
|
35 |
+
index = faiss.IndexFlatL2(dimension)
|
36 |
+
index.add(embeddings)
|
37 |
+
|
38 |
+
# Save index and contexts
|
39 |
+
with open(INDEX_FILE, "wb") as f:
|
40 |
+
pickle.dump(index, f)
|
41 |
+
with open(DOCS_FILE, "wb") as f:
|
42 |
+
pickle.dump(contexts, f)
|
43 |
+
|
44 |
+
progress.progress(100)
|
45 |
+
return index, contexts
|
46 |
+
|
47 |
+
# Load existing index or build
|
48 |
+
if os.path.exists(INDEX_FILE) and os.path.exists(DOCS_FILE):
|
49 |
+
with open(INDEX_FILE, "rb") as f:
|
50 |
+
faiss_index = pickle.load(f)
|
51 |
+
with open(DOCS_FILE, "rb") as f:
|
52 |
+
all_contexts = pickle.load(f)
|
53 |
+
else:
|
54 |
+
faiss_index, all_contexts = setup_database()
|
55 |
+
|
56 |
+
# Sample questions
|
57 |
+
sample_questions = [
|
58 |
+
"What is the role of Falcon RefinedWeb in this dataset?",
|
59 |
+
"How can retrieval improve language generation?",
|
60 |
+
"Explain the purpose of the RAG dataset."
|
61 |
]
|
62 |
|
63 |
+
st.subheader("Ask a question based on the dataset:")
|
64 |
+
question = st.text_input("Your question", value=sample_questions[0])
|
65 |
+
|
66 |
if st.button("Ask"):
|
67 |
+
with st.spinner("Retrieving relevant context and generating answer..."):
|
68 |
+
embedder = SentenceTransformer(MODEL_NAME)
|
69 |
+
question_embedding = embedder.encode([question])
|
70 |
+
D, I = faiss_index.search(question_embedding, k=1)
|
71 |
+
|
72 |
+
retrieved_context = all_contexts[I[0][0]]
|
73 |
+
prompt = f"Context: {retrieved_context}\n\nQuestion: {question}\n\nAnswer:"
|
74 |
+
|
75 |
+
response = client.chat.completions.create(
|
76 |
+
messages=[{"role": "user", "content": prompt}],
|
77 |
+
model="llama-3-70b-8192"
|
78 |
+
)
|
79 |
+
|
80 |
+
answer = response.choices[0].message.content
|
81 |
+
st.success("Answer:")
|
82 |
+
st.write(answer)
|
83 |
+
|
84 |
+
with st.expander("Retrieved Context"):
|
85 |
+
st.markdown(retrieved_context)
|