File size: 1,235 Bytes
6f5969f | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 | import gradio as gr
import pickle, random
from sentence_transformers import SentenceTransformer
# Load model (must be in the same folder)
with open("intent_model.pkl", "rb") as f:
data = pickle.load(f)
clf = data["classifier"]
id2label = data["id2label"]
embedder = SentenceTransformer(data["embed_model"])
intents_meta = data["intents_meta"]
def predict_intent(user_input):
"""Predict intent and return formatted response."""
if not user_input.strip():
return "Please enter a command."
emb = embedder.encode([user_input])
pred = clf.predict(emb)[0]
intent = id2label[pred]
meta = intents_meta[intent]
response = random.choice(meta["responses"])
action = meta["action"]
return f"🧠 Intent: {intent}\n💬 Response: {response}\n⚙️ Action: {action}"
# Gradio interface
demo = gr.Interface(
fn=predict_intent,
inputs=gr.Textbox(label="Enter your command", placeholder="e.g. restart pc, open gmail"),
outputs=gr.Textbox(label="Jarvis Response"),
title="🧠 Jarvis Intent Classifier",
description="Lightweight intent classification model that detects system commands and returns appropriate responses & actions."
)
if __name__ == "__main__":
demo.launch()
|