feat(nutrisight): add a demo that can be run as-it

nutrisight/README.md

@@ -11,6 +11,42 @@ Then, we analyzed where the model predictions differed from the annotations, to
 
 The model is also available on [Hugging Face](https://huggingface.co/openfoodfacts/nutrition-extractor).
 
+You can easily test the model by running the `demo.py` script. The requirements are described as metadata at the top of the file, and
+the script can be run using [uv](https://docs.astral.sh/uv/) in a single step:
+
+```bash
+$ uv run demo.py https://images.openfoodfacts.org/images/products/406/826/203/8564/2.jpg
+
+Reading inline script metadata from `demo.py`
+Running nutrition extraction model on:
+image URL: https://images.openfoodfacts.org/images/products/406/826/203/8564/2.jpg
+OCR URL: https://images.openfoodfacts.org/images/products/406/826/203/8564/2.json
+
+Nutrient predictions
+--------------------
+serving_size              45 g       (score: 0.999)
+energy-kj_100g            1928 kj    (score: 0.999)
+energy-kj_serving         868 kj     (score: 0.998)
+energy-kcal_100g          460 kcal   (score: 0.999)
+energy-kcal_serving       207 kcal   (score: 0.998)
+fat_100g                  18 g       (score: 0.999)
+fat_serving               8.0 g      (score: 0.998)
+saturated-fat_100g        5.3 g      (score: 0.999)
+saturated-fat_serving     2.4 g      (score: 0.998)
+carbohydrates_100g        62 g       (score: 0.999)
+carbohydrates_serving     28 g       (score: 0.998)
+sugars_100g               22 g       (score: 0.999)
+sugars_serving            9.8 g      (score: 0.998)
+fiber_100g                7.3 g      (score: 0.998)
+fiber_serving             3.3 g      (score: 0.997)
+proteins_100g             9.3 g      (score: 0.999)
+proteins_serving          4.2 g      (score: 0.997)
+salt_100g                 0.08 g     (score: 0.999)
+salt_serving              0.04 g     (score: 0.998)
+```
+
+If you don't provide an Open Food Facts image, you should also provide the URL of the OCR result generated by Google Cloud Vision.
+
 ## Dataset management
 
 Every script related to dataset management and generation is located in the `dataset` directory: