Embedding dimensionality is the number of floating-point values in each vector. It is an index-level choice: higher dimensions can capture more nuanced semantic relationships, but require more storage and slow similarity calculations. Set the embedding model and dimensions when creating an index; keep its vectors comparable by using a consistent model. section 5.3.6 section 5.2.2
Treat an embedding configuration change as an index migration:
- Create a new index with the desired configuration and re-ingest documents. section 5.2.2
- Validate retrieval quality, switch traffic after validation, and retain the old index as a rollback option. section 5.2.2
Listing 5.1 Index configuration dataclass
@dataclass
class IndexConfig:
name: str #A
embedding_model: str =
➥"text-embedding-3-small" #B
embedding_dimensions: int = 1536 #B
chunking_strategy: str = "fixed" #C
chunk_size: int = 512 #C
chunk_overlap: int = 50 #C
metadata_schema: Optional[Dict] = None #DThe index configuration dataclass makes the dimension an explicit setting, with a default of 1536. section 5.2.2










