Crate serdes_ai_embeddings 

serdes-ai-embeddings

Embedding models and vector operations for serdes-ai.

This crate provides infrastructure for generating embeddings from text and performing vector similarity operations.

Core Concepts

• EmbeddingModel: Trait for embedding model implementations
• Embedding: Vector representation with metadata
• Embedder: High-level interface for embeddings
• Similarity functions: Cosine, dot product, Euclidean distance

Feature Flags

• openai (default): OpenAI embedding models
• cohere: Cohere embedding models
• voyage: Voyage AI embeddings
• ollama: Local Ollama embeddings
• full: All providers

Example

use serdes_ai_embeddings::{Embedder, EmbeddingModel};

// Using the high-level interface
let embedder = Embedder::from_env("openai:text-embedding-3-small")?;

// Single embedding
let result = embedder.embed_query("Hello, world!").await?;
let embedding = result.embedding().unwrap();

// Batch embeddings
let result = embedder.embed_documents(vec![
    "First document".into(),
    "Second document".into(),
]).await?;

// Similarity
let similarity = result.embeddings[0].cosine_similarity(&result.embeddings[1]);

Direct Model Usage

use serdes_ai_embeddings::OpenAIEmbeddingModel;

let model = OpenAIEmbeddingModel::from_env("text-embedding-3-small")?;
let result = model.embed_query("Hello!").await?;

Re-exports

pub use embedder::infer_embedding_model;

pub use embedder::Embedder;

pub use embedder::EmbedderBuilder;

pub use embedding::Embedding;

pub use embedding::EmbeddingBatch;

pub use error::EmbeddingError;

pub use error::EmbeddingResult;

pub use model::BoxedEmbeddingModel;

pub use model::EmbedInput;

pub use model::EmbeddingModel;

pub use model::EmbeddingOutput;

pub use model::EmbeddingSettings;

pub use model::EncodingFormat;

pub use model::InputType;

pub use model::TruncationMode;

pub use similarity::angular_distance;

pub use similarity::centroid;

pub use similarity::cosine_similarity;

pub use similarity::dot_product;

pub use similarity::euclidean_distance;

pub use similarity::manhattan_distance;

pub use similarity::normalize;

pub use similarity::pairwise_cosine;

pub use similarity::top_k_similar;

pub use similarity::weighted_average;

pub use openai::OpenAIEmbeddingModel;openai

Modules

embedder

High-level embedder interface.

embedding

Embedding types and batch operations.

error

Embedding errors.

model

Embedding model trait and types.

openaiopenai

OpenAI embedding model implementation.

prelude

Prelude for common imports.

similarity

Similarity and distance functions for embeddings.