Struct SemanticConfig 

pub struct SemanticConfig {
    pub embedding_model: String,
    pub embedding_model_path: Option<String>,
    pub top_k: usize,
    pub embeddings_endpoint: Option<String>,
    pub embeddings_api: Option<BackendKind>,
    pub on_embed_failure: OnEmbedFailure,
}

[context.semantic] — the embedding RAG-for-code feature’s settings (Step 26.5.4, #582).

Fields

embedding_model: String

Embedding model used to index the repo + embed queries. Default nomic-embed-text (the HTTP path). The model must exist on the embeddings endpoint (see embeddings_endpoint); when it can’t be reached the feature follows on_embed_failure.

For the embedded backend (embeddings_api = "embedded", #720) this is only a label — the model is loaded from embedding_model_path — and it should name a candle-clean standard-BERT model (e.g. bge-small-en-v1.5), NOT nomic-embed-text, which candle 0.8 cannot load.

embedding_model_path: Option<String>

Local model directory for the embedded embedder (#720): a candle-clean standard-BERT model dir holding config.json + tokenizer.json + model.safetensors (e.g. a fetched BAAI/bge-small-en-v1.5). None (default) ⇒ the embedded path can’t load and reports a clear error. When embeddings_api and embeddings_endpoint are unset, a configured path selects embedded embeddings automatically. Ignored by explicit HTTP embeddings targets. Mirrors the summarizer’s model_path.

top_k: usize

How many code chunks to retrieve per turn. Default 5.

embeddings_endpoint: Option<String>

Dedicated endpoint that serves embeddings (e.g. an Ollama http://host:11434). None (default) leaves semantic retrieval on the embedded path unless embeddings_api explicitly selects an HTTP protocol. Set this to a real embeddings host when remote/vector-server embeddings are a deliberate performance choice.

embeddings_api: Option<BackendKind>

Wire protocol of embeddings_endpoint — ollama (/api/embeddings) or openai (/v1/embeddings). embedded selects the in-process embedder. None (default) selects embedded embeddings when embeddings_endpoint is also unset; with an explicit endpoint, None assumes ollama.

on_embed_failure: OnEmbedFailure

What to do when embedding fails structurally (wrong endpoint / model absent): disable (default) stops indexing after the first failure with one actionable message; warn logs per-chunk and keeps trying.

Trait Implementations

impl Clone for SemanticConfig

fn clone(&self) -> SemanticConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SemanticConfig

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

Formats the value using the given formatter.

impl Default for SemanticConfig

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for SemanticConfig

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Eq for SemanticConfig

impl PartialEq for SemanticConfig

fn eq(&self, other: &SemanticConfig) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Serialize for SemanticConfig

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

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for SemanticConfig