Update README.md

README.md

@@ -1,16 +1,32 @@
+---
+title: Pricing Optimization API
+emoji: 🚗
+colorFrom: blue
+colorTo: green
+sdk: docker
+app_port: 7860
+pinned: false
+license: mit
+---
+
 # Pricing Optimization API
 
-Production-ready FastAPI for `/predict` with built-in `/docs` (Swagger UI).  
-Just drop your trained bundle `RF_model.joblib` (containing `{"model": pipeline, "feature_names": [...]}`) into this folder and deploy.
+This is a production-ready **FastAPI** service that serves your **scikit-learn Pipeline**.
+It exposes:
+- `POST /predict` — batch predictions
+- `GET /docs` — interactive Swagger UI
+- `GET /` — simple health check
 
-## Endpoints
-- `GET /` — health check
-- `POST /predict` — body: `{"input": [[f1, f2, ..., fN], ...]}` → returns `{"prediction": [..]}`
-- `GET /docs` — interactive API docs
+## Why this setup?
+- The saved artifact `RF_model.joblib` contains both preprocessing and the model, so serving is consistent with training.
+- The API accepts mixed types (strings/ints/bools/floats) and converts them into a **pandas DataFrame** with the **exact** column names used at training, allowing `ColumnTransformer` to work reliably.
 
-## Input format
-The model expects **N features** in a specific order (printed to logs on startup). For mixed types (categorical + numeric + booleans), send them exactly as trained. Example:
+## Expected input format
+- Body: `{"input": [[f1, f2, ..., fN], ...]}` — 2D list (batch of rows).
+- **Order matters**: the order must match `feature_names` saved in your bundle.
+- Booleans must be JSON booleans (`true`/`false`) — not strings.
 
+**Example**
 ```json
 {
   "input": [[