Create app.py

app.py

@@ -0,0 +1,51 @@
+import gradio as gr
+from sentence_transformers import SentenceTransformer, util
+import networkx as nx
+
+# Load pre-trained model (runs only once at startup)
+model = SentenceTransformer("sentence-transformers/paraphrase-mpnet-base-v2")
+
+def group_duplicates_api(complaints, threshold=0.7):
+    """
+    Input:
+      complaints: list of complaint strings
+      threshold: float between 0-1 (default 0.7)
+    Output:
+      List of groups, where each group is a list of complaint texts
+    """
+    if not complaints:
+        return []
+
+    embeddings = model.encode(complaints, convert_to_tensor=True)
+    cosine_scores = util.pytorch_cos_sim(embeddings, embeddings)
+
+    G = nx.Graph()
+    G.add_nodes_from(range(len(complaints)))
+
+    for i in range(len(complaints)):
+        for j in range(i + 1, len(complaints)):
+            if cosine_scores[i][j].item() >= threshold:
+                G.add_edge(i, j)
+
+    duplicate_groups = list(nx.connected_components(G))
+    results = []
+    for group in duplicate_groups:
+        group_texts = [complaints[idx] for idx in group]
+        results.append(group_texts)
+
+    return results
+
+# Gradio interface
+demo = gr.Interface(
+    fn=group_duplicates_api,
+    inputs=[
+        gr.Textbox(lines=10, placeholder="Enter complaints separated by newline", label="Complaints"),
+        gr.Slider(0.5, 0.95, value=0.7, step=0.01, label="Similarity Threshold")
+    ],
+    outputs=gr.JSON(label="Duplicate Groups"),
+    title="Duplicate Complaint Grouping API",
+    description="Paste multiple complaints (one per line) and get grouped duplicates."
+)
+
+if __name__ == "__main__":
+    demo.launch()