Module embeddings 

Embedding module for semantic search.

Provides embedding generation using HTTP-based providers:

• Ollama (local) - Recommended for development
• HuggingFace (cloud) - Requires API token

Architecture

┌──────────────────┐
│  CLI Commands    │
│ (status/backfill)│
└────────┬─────────┘
         │
         ▼
┌─────────────────┐
│    Factory      │  ← Auto-detects available provider
└────────┬────────┘
         │
    ┌────┴────┐
    ▼         ▼
┌───────┐ ┌───────────┐
│Ollama │ │HuggingFace│
└───────┘ └───────────┘
    │         │
    ▼         ▼
  HTTP      HTTP
localhost  API

Configuration

Settings are loaded from ~/.savecontext/config.json to maintain compatibility with the TypeScript MCP server.

Environment variables take precedence:

• OLLAMA_ENDPOINT - Ollama server URL (default: http://localhost:11434)
• OLLAMA_MODEL - Embedding model (default: nomic-embed-text)
• HF_TOKEN - HuggingFace API token
• HF_MODEL - HuggingFace model (default: sentence-transformers/all-MiniLM-L6-v2)
• SAVECONTEXT_EMBEDDINGS_ENABLED - Enable/disable embeddings (default: true)

Usage

use sc::embeddings::{create_embedding_provider, detect_available_providers};

// Detect available providers
let detection = detect_available_providers().await;
println!("Available: {:?}", detection.available);

// Create provider (auto-detects)
if let Some(provider) = create_embedding_provider().await {
    let info = provider.info();
    println!("Using {} ({})", info.name, info.model);

    let embedding = provider.generate_embedding("Hello world").await?;
    println!("Dimensions: {}", embedding.len());
}

Re-exports

pub use config::get_embedding_settings;

pub use config::is_embeddings_enabled;

pub use config::resolve_hf_model;

pub use config::resolve_hf_token;

pub use config::resolve_ollama_endpoint;

pub use config::resolve_ollama_model;

pub use config::reset_embedding_settings;

pub use config::save_embedding_settings;

pub use factory::create_embedding_provider;

pub use factory::create_huggingface_provider;

pub use factory::create_ollama_provider;

pub use factory::detect_available_providers;

pub use factory::ProviderDetection;

pub use huggingface::HuggingFaceProvider;

pub use model2vec::Model2VecProvider;

pub use ollama::OllamaProvider;

pub use provider::BoxedProvider;

pub use provider::EmbeddingProvider;

pub use types::EmbeddingProviderType;

pub use types::EmbeddingResult;

pub use types::EmbeddingSettings;

pub use types::ModelConfig;

pub use types::ProviderInfo;

pub use types::SaveContextConfig;

pub use types::SearchMode;

pub use types::TieredEmbeddingSettings;

pub use types::model2vec_models;

pub use chunking::chunk_text;

pub use chunking::prepare_item_text;

pub use chunking::ChunkConfig;

pub use chunking::TextChunk;

Modules

chunking

Text chunking for embeddings.

config

Embedding configuration management.

factory

Embedding provider factory.

huggingface

HuggingFace Inference API embedding provider.

model2vec

Model2Vec embedding provider.

ollama

Ollama embedding provider.

provider

Embedding provider trait.

types

Embedding types and configuration.