René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"to_ray_dataset as a first citizen method for retrieval_job","articleBody":"we probably can make it a first citizen method for retrieval_job in future\n\n_Originally posted by @HaoXuAI in https://github.com/feast-dev/feast/pull/5526#discussion_r2240629274_\n\n```\n\nfrom feast import FeatureStore\n\nstore = FeatureStore(\".\")\n\n# Get historical features as Ray Dataset\nds = store.get_historical_features(\n    entity_df=entity_df,\n    features=[\"embeddings:embedding\"],\n).to_ray_dataset()  # ← Proposed new method\n\n# Now use Ray Data API directly\nds = ds.map_batches(custom_processing)\nds = ds.repartition(100)\npredictions = ds.map_batches(ModelInference, num_gpus=1)\n```\n\n```\nclass FeatureStore:\n    \n    def to_ray_dataset(\n        self,\n        source: Optional[DataSource] = None,\n        feature_view: Optional[FeatureView] = None,\n        start_date: Optional[datetime] = None,\n        end_date: Optional[datetime] = None,\n    ) -\u003e ray.data.Dataset:\n        \"\"\"Convert Feast data source or feature view to Ray Dataset.\"\"\"\n        ...\n    \nclass RetrievalJob:\n    \n    def to_ray_dataset(self) -\u003e ray.data.Dataset:\n        \"\"\"Convert retrieval result to Ray Dataset instead of Pandas.\"\"\"\n        ...\n```\n            ","author":{"url":"https://github.com/ntkathole","@type":"Person","name":"ntkathole"},"datePublished":"2025-08-16T04:29:28.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":0},"url":"https://github.com/5568/feast/issues/5568"}