For Japanese version, see IEmbeddingProvider-jp.md.

IEmbeddingProvider (Interface Specification)

1. Responsibility Boundary

IEmbeddingProvider converts text into semantic vectors and supplies numeric foundations for semantic-space operations in Phase 1.

• Role:

Provide single/batch embedding generation and vector-dimension metadata.

• Non-role:

Vector persistence, indexing, nearest-neighbor search, and Core knowledge retrieval are out of scope.

2. Contract Signature

namespace AIKernel.Abstractions.Providers;\n\npublic interface ITextEmbeddingProvider\n{\n    Task<float[]> EmbedAsync(string text, CancellationToken cancellationToken = default);\n}\n\npublic interface IBatchEmbeddingProvider\n{\n    Task<IReadOnlyList<float[]>> EmbedBatchAsync(IReadOnlyList<string> texts, CancellationToken cancellationToken = default);\n}\n\npublic interface IEmbeddingDimensionProvider\n{\n    int GetDimension();\n}\n\npublic interface IEmbeddingProvider :\n    IProvider,\n    ITextEmbeddingProvider,\n    IBatchEmbeddingProvider,\n    IEmbeddingDimensionProvider\n{\n}

3. Related Use Cases

• UC-05 Material relevance evaluation:

Uses embeddings for similarity-based material filtering.

• UC-11 RAG:

Projects queries and material chunks into a common vector space when a provider or pipeline strategy performs material retrieval.

• Phase 1 Query Processing:

Supports optional semantic projection after IQueryDecomposer has produced QueryPart units.

4. Governance & Determinism

• Model stability:

For the same ProviderId and configuration, embedding dimensionality/space behavior should not drift silently.

• Fail-Closed:

Input-limit violations or dimension mismatches should fail explicitly instead of returning partial semantics.

• RAG boundary:

Embedding support does not make Core a retrieval engine; it only exposes semantic projection capability.

5. Implementation Notes

• Dimension consistency:

Ensure GetDimension() always matches produced vector lengths.

• Batch throttling:

EmbedBatchAsync should consider provider rate limits and apply chunking when required.

Changelog

• v0.0.0 / v0.0.0.0: Initial draft
• v0.0.1 (2026-05-06): Version upgrade aligned with documentation guidelines
• v0.0.1 (2026-05-09): Aligned embedding docs with split capabilities and Phase 1 Query Processing