Update app.py

app.py

@@ -1,38 +1,57 @@
-import gradio as gr
+# app.py
+import os
 import pandas as pd
+import numpy as np
 from sentence_transformers import SentenceTransformer
 import faiss
+import gradio as gr
+
+# --- 1) Load the CSV you just uploaded ---
+df = pd.read_csv("quotes.csv")
+corpus = df["text"].astype(str).tolist()
+
+# --- 2) Prepare embedding model ---
+MODEL_NAME = "all-MiniLM-L6-v2"
+model = SentenceTransformer(MODEL_NAME)
 
-# 1. Dataset — replace or load from CSV
-texts = [
-    "Creativity is intelligence having fun.",
-    "Simplicity is the ultimate sophistication.",
-    "Innovation distinguishes between a leader and a follower.",
-    # Add more...
-]
-df = pd.DataFrame({"text": texts})
+# --- 3) Build or load embeddings + FAISS index (cached if already computed) ---
+EMB_FILE = "embeddings.npy"
+IDX_FILE = "index.faiss"
 
-# 2. Embeddings + FAISS index
-model = SentenceTransformer("all-MiniLM-L6-v2")
-emb = model.encode(df["text"].tolist(), convert_to_numpy=True).astype("float32")
-faiss.normalize_L2(emb)
-index = faiss.IndexFlatIP(emb.shape[1])
-index.add(emb)
+if os.path.exists(EMB_FILE) and os.path.exists(IDX_FILE):
+    corpus_embeddings = np.load(EMB_FILE)
+    index = faiss.read_index(IDX_FILE)
+else:
+    corpus_embeddings = model.encode(corpus, convert_to_numpy=True).astype("float32")
+    # normalize for cosine
+    faiss.normalize_L2(corpus_embeddings)
+    index = faiss.IndexFlatIP(corpus_embeddings.shape[1])
+    index.add(corpus_embeddings)
+    np.save(EMB_FILE, corpus_embeddings)
+    faiss.write_index(index, IDX_FILE)
 
-# 3. Search function
-def search(query):
+# --- 4) Search function ---
+def get_top3(query):
+    if not query or query.strip()=="":
+        return "Type something to see 3 similar quotes."
     qv = model.encode([query], convert_to_numpy=True).astype("float32")
     faiss.normalize_L2(qv)
-    D, I = index.search(qv, 3)  # top 3
-    return "\n".join([df.iloc[i]["text"] for i in I[0]])
+    D, I = index.search(qv, 3)
+    res = []
+    for idx in I[0]:
+        res.append(corpus[idx])
+    return "\n\n".join([f"{i+1}. {r}" for i,r in enumerate(res)])
 
-# 4. Gradio UI
+# --- 5) Gradio UI ---
+examples = [["creativity"], ["return process"], ["make returns effortless"]]
 demo = gr.Interface(
-    fn=search,
-    inputs=gr.Textbox(lines=2, placeholder="Type something..."),
-    outputs="text",
-    title="Mini Quote Finder",
-    description="Type a phrase, get similar quotes."
+    fn=get_top3,
+    inputs=gr.Textbox(lines=2, placeholder="Type a phrase (e.g., 'make returns easy')"),
+    outputs=gr.Textbox(label="Top 3 similar quotes"),
+    title="Quote Finder (semantic search)",
+    description="Type something and get 3 similar quotes from the dataset.",
+    examples=examples
 )
 
-demo.launch()
+if __name__ == "__main__":
+    demo.launch()