Create app.py

app.py

@@ -0,0 +1,27 @@
+import gradio as gr
+from transformers import AutoTokenizer, AutoModel
+import torch
+
+# Load public model
+model_name = "intfloat/multilingual-e5-small"
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModel.from_pretrained(model_name)
+
+# Inference function
+def get_embedding(text):
+    # E5 models expect: "query: your text here"
+    encoded_input = tokenizer("query: " + text, return_tensors='pt')
+    with torch.no_grad():
+        model_output = model(**encoded_input)
+
+    embeddings = model_output.last_hidden_state[:, 0]  # CLS token
+    normed = torch.nn.functional.normalize(embeddings, p=2, dim=1)
+    return normed[0].tolist()  # return list
+
+# Gradio UI
+iface = gr.Interface(fn=get_embedding,
+                     inputs=gr.Textbox(label="Enter text"),
+                     outputs=gr.Textbox(label="Embedding"),
+                     title="Text Embedder")
+
+iface.launch()