Struct EmbeddingConfig 

pub struct EmbeddingConfig {
    pub provider: EmbeddingProvider,
    pub model: String,
    pub dimensions: usize,
    pub endpoint: Option<String>,
    pub api_key: Option<String>,
    pub api_version: Option<String>,
    pub max_completion_tokens: usize,
    pub batch_size: usize,
    pub mock: bool,
    pub mock_mode: MockVectorMode,
    pub onnx: OnnxEmbeddingConfig,
    pub huggingface_tokenizer: Option<String>,
}

Unified embedding configuration.

Provider-agnostic; holds fields for all supported backends. Load from environment variables via EmbeddingConfig::from_env, or construct programmatically and pass to EmbeddingConfig::create_engine.

Environment variables (match Python SDK names):

• EMBEDDING_PROVIDER — backend selection (default: openai; onnx on Android)
• MOCK_EMBEDDING — set to true/1/yes to force mock mode
• EMBEDDING_MODEL — model identifier
• EMBEDDING_DIMENSIONS — vector size
• EMBEDDING_ENDPOINT — API endpoint URL
• EMBEDDING_API_KEY — API key (fallback: LLM_API_KEY)
• EMBEDDING_API_VERSION — API version string
• EMBEDDING_MAX_COMPLETION_TOKENS — maximum tokens (default: 8191)
• EMBEDDING_BATCH_SIZE — texts per embedding request (default: 36)
• EMBEDDING_ONNX_BATCH_SIZE — ONNX inference batch size (default: 32; onnx feature only)
• HUGGINGFACE_TOKENIZER — HuggingFace tokenizer identifier

Fields

provider: EmbeddingProvider

Which backend to use for embedding generation.

model: String

Model identifier. For ONNX this is informational; for API providers this is sent in the request body. Default depends on provider (BGE-Small-v1.5 for ONNX, empty for others).

dimensions: usize

Embedding vector dimensionality. Must match the model output.

endpoint: Option<String>

API endpoint URL (used by OpenAI-compatible and Ollama providers).

api_key: Option<String>

API key. Reads EMBEDDING_API_KEY first, falls back to LLM_API_KEY.

api_version: Option<String>

API version string (e.g. “2023-05-15” for Azure OpenAI).

max_completion_tokens: usize

Maximum tokens for completion requests (default: 8191).

batch_size: usize

Number of texts to send in a single embedding request (default: 36).

Matches the Python SDK and stays within the small client-batch limits of the self-hosted servers this adapter targets (e.g. TEI defaults to 32). Raise it via EMBEDDING_BATCH_SIZE for providers that accept larger batches. For the OpenAI-compatible engine, up to MAX_CONCURRENT_BATCHES sub-batches are also dispatched concurrently.

mock: bool

If true, use mock embeddings regardless of provider. Overrides provider to Mock. Set via MOCK_EMBEDDING=true.

mock_mode: MockVectorMode

How the mock engine generates vectors when provider is Mock. Defaults to MockVectorMode::Zero. Set via MOCK_EMBEDDING=deterministic to derive content-stable vectors from sha256(text).

onnx: OnnxEmbeddingConfig

ONNX-specific configuration. Only consulted when provider is Onnx or Fastembed.

huggingface_tokenizer: Option<String>

HuggingFace tokenizer identifier for chunking token counting. When set, used by HuggingFaceTokenCounter in the chunking crate.

Implementations

impl EmbeddingConfig

pub fn from_env() -> Self

Load configuration from environment variables.

Reads the same env var names as the Python SDK so that a shared .env file works across both implementations without modification.

pub fn effective_provider(&self) -> EmbeddingProvider

Returns the effective provider, substituting Mock when self.mock is true.

pub async fn create_engine(&self) -> EmbeddingResult<Arc<dyn EmbeddingEngine>>

Create an embedding engine based on this configuration.

Dispatches to the appropriate engine implementation based on EmbeddingConfig::effective_provider. Providers not yet implemented return [EmbeddingError::NotImplemented].

Trait Implementations

impl Clone for EmbeddingConfig

fn clone(&self) -> EmbeddingConfig

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for EmbeddingConfig

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for EmbeddingConfig

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for EmbeddingConfig

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for EmbeddingConfig

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.