Update app.py

app.py

@@ -1,6 +1,5 @@
-from fastapi import FastAPI
-from pydantic import BaseModel
-from typing import List
+# app.py
+import gradio as gr
 import torch
 from transformers import (
     AutoTokenizer,
@@ -8,8 +7,6 @@ from transformers import (
     XLMRobertaForSequenceClassification,
 )
 
-app = FastAPI(title="Unified NLP API")
-
 # =====================
 # Agreement (MNLI)
 # =====================
@@ -32,7 +29,6 @@ def check_agreement(msg1: str, msg2: str) -> float:
     probs = torch.softmax(logits, dim=-1)[0]
     return round((probs[2] - probs[0]).item(), 2)  # entailment - contradiction
 
-
 # =====================
 # Sentiment
 # =====================
@@ -56,17 +52,14 @@ def analyze_sentiment(text: str) -> float:
     stars = torch.argmax(probs, dim=-1).item() + 1
     return round((stars - 3) * 2.5, 2)  # -5 .. +5
 
-
 # =====================
 # Multilabel classifier
 # =====================
 CLASSIFIER_MODEL = "xlm-roberta-base"
-
 CATEGORIES = [
     "politique", "woke", "racism", "crime",
     "police_abuse", "corruption", "hate_speech", "activism"
 ]
-
 clf_tokenizer = None
 clf_model = None
 
@@ -76,11 +69,12 @@ def load_classifier():
         clf_tokenizer = AutoTokenizer.from_pretrained(CLASSIFIER_MODEL)
         clf_model = XLMRobertaForSequenceClassification.from_pretrained(
             CLASSIFIER_MODEL,
-            num_labels=len(CATEGORIES)
+            num_labels=len(CATEGORIES),
+            problem_type="multi_label_classification"
         )
         clf_model.eval()
 
-def classify_message(text: str) -> List[str]:
+def classify_message(text: str) -> list:
     load_classifier()
     inputs = clf_tokenizer(text, return_tensors="pt", truncation=True)
     with torch.no_grad():
@@ -89,42 +83,31 @@ def classify_message(text: str) -> List[str]:
     labels = [CATEGORIES[i] for i, p in enumerate(probs) if p > 0.5]
     return labels or ["neutral"]
 
-
-# =====================
-# API schemas
-# =====================
-class AgreementRequest(BaseModel):
-    msg1: str
-    msg2: str
-
-class TextRequest(BaseModel):
-    text: str
-
-
 # =====================
-# Endpoints
+# Gradio interfaces
 # =====================
-@app.post("/agreement")
-def agreement(req: AgreementRequest):
-    return {"agreement_score": check_agreement(req.msg1, req.msg2)}
-
-@app.post("/sentiment")
-def sentiment(req: TextRequest):
-    return {"sentiment_score": analyze_sentiment(req.text)}
-
-@app.post("/classify")
-def classify(req: TextRequest):
-    return {"categories": classify_message(req.text)}
-
-@app.get("/")
-def root():
-    return {
-        "status": "ok",
-        "endpoints": {
-            "POST /sentiment": "sentiment analysis",
-            "POST /agreement": "text agreement",
-            "POST /classify": "multilabel classification",
-            "GET /docs": "swagger UI"
-        }
-    }
 
+with gr.Blocks(title="Unified NLP API") as demo:
+
+    gr.Markdown("## 📈 Unified NLP API")
+    
+    with gr.Tab("Agreement"):
+        msg1 = gr.Textbox(label="Message 1")
+        msg2 = gr.Textbox(label="Message 2")
+        btn_agree = gr.Button("Check Agreement")
+        out_agree = gr.Number(label="Agreement Score")
+        btn_agree.click(fn=check_agreement, inputs=[msg1, msg2], outputs=out_agree)
+    
+    with gr.Tab("Sentiment"):
+        text_sent = gr.Textbox(label="Text")
+        btn_sent = gr.Button("Analyze Sentiment")
+        out_sent = gr.Number(label="Sentiment Score (-5 to +5)")
+        btn_sent.click(fn=analyze_sentiment, inputs=text_sent, outputs=out_sent)
+    
+    with gr.Tab("Multilabel Classification"):
+        text_clf = gr.Textbox(label="Text")
+        btn_clf = gr.Button("Classify")
+        out_clf = gr.Label(label="Categories")
+        btn_clf.click(fn=classify_message, inputs=text_clf, outputs=out_clf)
+
+demo.launch()