EmbeddingChunk

EmbeddingChunk represents an embedding vector returned by the embedding adapter.

Overview

The EmbeddingChunk class represents a single embedding vector returned by an embedding adapter. It provides a standardized structure for embedding results that maintains compatibility with OpenAI's embedding response format while adding type safety.

Class Definition

class EmbeddingChunk(BaseModel):
    embedding: Sequence[float]
    index: int

Attributes

embedding

• Type: Sequence[float]
• Description: The embedding vector as a sequence of floating-point numbers. This represents the semantic representation of the input text in vector space.

index

• Type: int
• Description: The original index of the corresponding text in the input sequence. This allows mapping embeddings back to their source texts when processing multiple inputs.

Usage Example

from amrita_core.types import EmbeddingChunk

# Create an embedding chunk
chunk = EmbeddingChunk(embedding=[0.1, -0.5, 0.8, 0.3], index=0)

print(f"Vector: {chunk.embedding}")
print(f"Original index: {chunk.index}")

# When processing multiple texts
texts = ["Hello", "World"]
embeddings: list[EmbeddingChunk] = await call_completion(
    preset=embedding_preset, messages=texts
)

for chunk in embeddings:
    print(f"Text '{texts[chunk.index]}' -> Embedding length: {len(chunk.embedding)}")

Related Components

• ModelAdapter.call_embed(): Method that returns EmbeddingChunk instances
• ModelAdapter: Base class for adapters, uses ADAPTER_TYPE literal type
• call_completion(): Function that handles embedding adapter calls