Deploy ExtraaLearn Flask backend

Dockerfile

@@ -0,0 +1,19 @@
+# Lightweight Python base image
+FROM python:3.12-slim
+
+# Hugging Face Docker spaces run the container as a non-root user with UID 1000
+RUN useradd -m -u 1000 user
+USER user
+ENV PATH="/home/user/.local/bin:$PATH"
+WORKDIR /home/user/app
+
+# Install dependencies first so Docker can cache this layer
+COPY --chown=user requirements.txt .
+RUN pip install --no-cache-dir --upgrade pip && pip install --no-cache-dir -r requirements.txt
+
+# Copy the API code and the serialized model
+COPY --chown=user . .
+
+# Serve the Flask app with gunicorn on port 7860 (the port Hugging Face exposes)
+EXPOSE 7860
+CMD ["gunicorn", "--workers", "2", "--bind", "0.0.0.0:7860", "--timeout", "120", "app:app"]

README.md

@@ -1,10 +1,15 @@
 ---
-title: ExtraaLearnLeadBackend
-emoji: 📚
-colorFrom: pink
-colorTo: pink
+title: ExtraaLearn Lead Conversion Backend
+emoji: 🎓
+colorFrom: blue
+colorTo: indigo
 sdk: docker
+app_port: 7860
 pinned: false
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+Flask REST API serving the ExtraaLearn lead-conversion model.
+
+* `GET /` - health check
+* `POST /v1/lead` - JSON body with one lead's features, returns prediction + conversion probability
+* `POST /v1/leadbatch` - multipart CSV upload (field `file`), returns one prediction per lead

app.py

@@ -0,0 +1,95 @@
+import joblib
+import pandas as pd
+from flask import Flask, jsonify, request
+
+# Initialise the Flask app and load the serialized pipeline (preprocessing + model) once at startup
+app = Flask("ExtraaLearn Lead Conversion Predictor")
+model = joblib.load("extraalearn_lead_model_v1_0.joblib")
+
+# Raw input features the pipeline expects, with their allowed values where categorical
+FEATURES = ["age", "current_occupation", "first_interaction", "profile_completed", "website_visits",
+            "time_spent_on_website", "page_views_per_visit", "last_activity", "print_media_type1",
+            "print_media_type2", "digital_media", "educational_channels", "referral"]
+NUMERIC = {"age": int, "website_visits": int, "time_spent_on_website": int, "page_views_per_visit": float}
+CATEGORIES = {
+    "current_occupation": ["Professional", "Unemployed", "Student"],
+    "first_interaction": ["Website", "Mobile App"],
+    "profile_completed": ["Low", "Medium", "High"],
+    "last_activity": ["Email Activity", "Phone Activity", "Website Activity"],
+    **{flag: ["Yes", "No"] for flag in ["print_media_type1", "print_media_type2", "digital_media",
+                                        "educational_channels", "referral"]},
+}
+
+
+def validate(record):
+    """Check that a lead record has every feature with a valid value; return a list of errors."""
+    errors = []
+    for f in FEATURES:
+        if f not in record:
+            errors.append(f"missing field '{f}'")
+        elif f in NUMERIC:
+            try:
+                if NUMERIC[f](record[f]) < 0:
+                    errors.append(f"'{f}' must be non-negative")
+            except (TypeError, ValueError):
+                errors.append(f"'{f}' must be numeric")
+        elif record[f] not in CATEGORIES[f]:
+            errors.append(f"'{f}' must be one of {CATEGORIES[f]}")
+    return errors
+
+
+def to_frame(records):
+    """Build a model-ready DataFrame with the right column order and numeric types."""
+    df = pd.DataFrame(records)[FEATURES]
+    for f, cast in NUMERIC.items():
+        df[f] = df[f].astype(cast)
+    return df
+
+
+@app.get("/")
+def home():
+    """Health check."""
+    return "Welcome to the ExtraaLearn Lead Conversion Prediction API!"
+
+
+@app.post("/v1/lead")
+def predict_lead():
+    """Predict the conversion of a single lead sent as a JSON object."""
+    record = request.get_json(silent=True)
+    if not isinstance(record, dict):
+        return jsonify({"error": "Request body must be a JSON object with the lead's features"}), 400
+    errors = validate(record)
+    if errors:
+        return jsonify({"error": "Invalid input", "details": errors}), 400
+
+    df = to_frame([record])
+    probability = float(model.predict_proba(df)[0, 1])
+    prediction = int(model.predict(df)[0])
+    return jsonify({
+        "prediction": prediction,
+        "prediction_label": "Likely to convert" if prediction == 1 else "Unlikely to convert",
+        "conversion_probability": round(probability, 4),
+    })
+
+
+@app.post("/v1/leadbatch")
+def predict_lead_batch():
+    """Predict the conversion of every lead in an uploaded CSV file (form field 'file')."""
+    if "file" not in request.files:
+        return jsonify({"error": "Upload a CSV file in the form field 'file'"}), 400
+    batch = pd.read_csv(request.files["file"])
+    missing = [f for f in FEATURES if f not in batch.columns]
+    if missing:
+        return jsonify({"error": f"CSV is missing columns: {missing}"}), 400
+
+    df = to_frame(batch)
+    ids = batch["ID"].astype(str).tolist() if "ID" in batch.columns else [str(i) for i in range(len(batch))]
+    results = [
+        {"ID": i, "prediction": int(p), "conversion_probability": round(float(pr), 4)}
+        for i, p, pr in zip(ids, model.predict(df), model.predict_proba(df)[:, 1])
+    ]
+    return jsonify(results)
+
+
+if __name__ == "__main__":
+    app.run(host="0.0.0.0", port=7860, debug=False)

extraalearn_lead_model_v1_0.joblib

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:928508ec696a1ca14e3204387b3283f134ab8d7f99433c33d256d95100640111
+size 132290

requirements.txt

@@ -0,0 +1,7 @@
+pandas==3.0.6
+numpy==2.5.3
+scikit-learn==1.9.1
+xgboost-cpu==3.4.1
+joblib==1.6.0
+flask==3.1.3
+gunicorn==26.2.0