'build'

.gitignore

@@ -0,0 +1,7 @@
+app.egg-info
+*.pyc
+.mypy_cache
+.coverage
+htmlcov
+.cache
+.venv

docker-compose.yml

@@ -0,0 +1,14 @@
+services:
+  bentoml-serve:
+    build: ./
+    container_name: bentoml_serve
+    ports:
+      - "3000:3000"
+    networks:
+      - traefik-public
+      - default
+
+networks:
+  traefik-public:
+    # Allow setting it to false for testing
+    external: false

dockerfile

@@ -0,0 +1,9 @@
+FROM python:3.10
+
+WORKDIR /app/
+
+COPY ./src /app/
+
+RUN pip install --no-cache-dir scikit-learn bentoml -i https://mirrors.aliyun.com/pypi/simple/
+
+CMD ["bentoml", "serve"]

src/service.py

@@ -0,0 +1,33 @@
+import bentoml
+import joblib
+
+my_image = bentoml.images.Image(python_version="3.10") \
+    .python_packages("scikit-learn", "numpy")
+
+@bentoml.service(
+    image=my_image
+)
+
+class NewsClassifierService:
+    model_logistic_path = "./models/logistic_model_0810.pkl"
+    model_vectorizer_path = "./models/vectorizer_0810.pkl"
+
+    def __init__(self) -> None:
+        self.model_logistic = joblib.load(self.model_logistic_path)
+        self.model_vectorizer = joblib.load(self.model_vectorizer_path)
+    
+    @bentoml.api
+    def classify(self,text: str) -> dict:
+        categories = ['Competition News','financial news','Medical news','sports news']
+        token_text =self.model_vectorizer.transform([text])
+        prediction = self.model_logistic.predict(token_text)
+        print(f"Prediction: {prediction}")
+        predict =  categories[prediction[0]]
+
+        json_response = {
+            "text": text,
+            "predict": predict
+        }
+
+        return json_response
+