Module embedding_provider 

Automatic Embedding Generation (Task 2)

This module provides colocated embedding resolution for text-to-vector conversion. It enables first-class text search by automatically generating embeddings.

Design

search_text(collection, text, k)
    │
    ▼
┌─────────────────┐
│ EmbeddingProvider │
│  ├─ LRU Cache    │
│  └─ ONNX Runtime │
└─────────────────┘
    │
    ▼
search_by_embedding(collection, embedding, k)

Providers

• LocalProvider: Uses FastEmbed/ONNX for offline embedding
• CachedProvider: LRU cache wrapper for any provider
• MockProvider: For testing

Complexity

• Embedding generation: O(n) where n = text length (transformer inference)
• Cache lookup: O(1) expected (hash-based LRU)
• Batch embedding: O(k) compute with ~O(1) ONNX session overhead

Structs

CacheStats

Cache statistics

CachedEmbeddingProvider

LRU-cached embedding provider wrapper

EmbeddingConfig

Configuration for embedding providers

EmbeddingVectorIndex

Vector index with automatic text embedding

LocalOnnxProvider

Local ONNX-based embedding provider

MockEmbeddingProvider

Mock embedding provider for testing

Enums

EmbeddingError

Error type for embedding operations

Traits

EmbeddingProvider

Embedding provider trait

Functions

create_embedding_index

Create an embedding-enabled vector index with mock provider

create_mock_provider

Create a cached mock embedding provider for testing

Type Aliases

EmbeddingResult

Result type for embedding operations