api.py

import os
import io
import json
from typing import Any, Dict, List

from flask import Flask, request, jsonify
from flask_cors import CORS
import pandas as pd

from loader import load_store_model

app = Flask(__name__)
CORS(app)

@app.route("/health", methods=["GET"])
def health():
    return jsonify({"status": "ok", "message": "SuperKart backend running"}), 200

def _predict_single(store_id: Any, features: Dict[str, Any]):
    model, meta = load_store_model(str(store_id))
    df = pd.DataFrame([features])
    yhat = model.predict(df)
    return float(yhat[0]), meta

@app.route("/predict", methods=["POST"])
#@app.post('/predict')
def predict():
    """POST JSON:
    {
      "store_id": "101",
      "features": { ... single row ... }
    }
    OR
    {
      "store_id": "101",
      "features_list": [ {...}, {...} ]   # multiple rows for same store
    }
    """
    try:
        payload = request.get_json(force=True, silent=False)
    except Exception as e:
        return jsonify({"error": f"Invalid JSON: {e}"}), 400

    if not payload:
        return jsonify({"error": "Empty payload"}), 400

    store_id = str(payload.get("store_id", "")).strip()
    if not store_id:
        return jsonify({"error": "Missing 'store_id'"}), 400

    try:
        model, meta = load_store_model(store_id)
    except FileNotFoundError as e:
        return jsonify({"error": str(e)}), 404

    if "features" in payload:
        df = pd.DataFrame([payload["features"]])
        yhat = model.predict(df)
        return jsonify({
            "store_id": store_id,
            "n_rows": 1,
            "predictions": [float(yhat[0])],
            "model": meta.get("model"),
            "metrics": meta.get("metrics", {}),
            "features_used": meta.get("features", [])
        }), 200

    elif "features_list" in payload:
        rows = payload["features_list"]
        if not isinstance(rows, list) or len(rows) == 0:
            return jsonify({"error": "'features_list' must be a non-empty list"}), 400
        df = pd.DataFrame(rows)
        yhat = model.predict(df)
        return jsonify({
            "store_id": store_id,
            "n_rows": len(df),
            "predictions": [float(v) for v in yhat],
            "model": meta.get("model"),
            "metrics": meta.get("metrics", {}),
            "features_used": meta.get("features", [])
        }), 200

    else:
        return jsonify({"error": "Provide either 'features' or 'features_list'"}), 400

@app.route("/predict_batch", methods=["POST"])
def predict_batch():
    """Multipart form with a CSV file:
    - expects a 'file' field
    - CSV must include a 'store_id' column and the necessary features.
    Will route each row to that store's model and return merged results.
    """
    if "file" not in request.files:
        return jsonify({"error": "No file uploaded with field name 'file'"}), 400
    f = request.files["file"]
    try:
        df = pd.read_csv(f)
    except Exception as e:
        return jsonify({"error": f"Failed to read CSV: {e}"}), 400

    if "store_id" not in df.columns:
        return jsonify({"error": "CSV must include 'store_id' column"}), 400

    preds = []
    errors = []
    # Simple cache for models during batch call
    cache = {}

    for idx, row in df.iterrows():
        sid = str(row["store_id"])
        feats = row.drop(labels=["store_id"]).to_dict()
        try:
            if sid not in cache:
                cache[sid] = load_store_model(sid)
            model, meta = cache[sid]
            yhat = model.predict(pd.DataFrame([feats]))[0]
            preds.append(float(yhat))
        except FileNotFoundError as e:
            preds.append(None)
            errors.append({"row": int(idx), "store_id": sid, "error": str(e)})
        except Exception as e:
            preds.append(None)
            errors.append({"row": int(idx), "store_id": sid, "error": f"{type(e).__name__}: {e}"})

    df_out = df.copy()
    df_out["predicted_sales"] = preds

    # Return as JSON (truncated) and CSV file bytes
    buf = io.StringIO()
    df_out.to_csv(buf, index=False)
    buf.seek(0)

    return jsonify({
        "rows": len(df_out),
        "errors": errors,
        "csv": buf.getvalue()
    }), 200

if __name__ == "__main__":
    # Local dev
    port = int(os.environ.get("PORT", 7860))
    app.run(host="0.0.0.0", port=port, debug=True)