All configuration structs derive Serialize and Deserialize, so they can be loaded from JSON, TOML, YAML files, or environment variables.

Module: magicaf_core::config

EmbeddingConfig

Configuration for the embedding service endpoint.

pub struct EmbeddingConfig {
    pub base_url: String,
    pub model_name: String,
    pub batch_size: usize,
    pub timeout_secs: u64,
    pub api_key: Option<String>,
}
FieldTypeDefaultDescription
base_urlString—Base URL of the embedding server (e.g., http://localhost:8080)
model_nameString—Model identifier sent in the request
batch_sizeusize32Maximum texts per batch request
timeout_secsu6430HTTP request timeout in seconds
api_keyOption<String>NoneOptional bearer token for authenticated endpoints

Example

let config = EmbeddingConfig {
    base_url: "http://localhost:8080".into(),
    model_name: "bge-large-en-v1.5".into(),
    batch_size: 32,
    timeout_secs: 30,
    api_key: None,
};

From JSON

{
  "base_url": "http://localhost:8080",
  "model_name": "bge-large-en-v1.5",
  "batch_size": 64,
  "timeout_secs": 15
}

VectorStoreConfig

Configuration for the vector database backend.

pub struct VectorStoreConfig {
    pub base_url: String,
    pub api_key: Option<String>,
    pub timeout_secs: u64,
}
FieldTypeDefaultDescription
base_urlString—Base URL of the vector store (e.g., http://localhost:6333)
api_keyOption<String>NoneOptional API key
timeout_secsu6430HTTP request timeout in seconds

LlmConfig

Configuration for the LLM service (OpenAI-compatible API).

pub struct LlmConfig {
    pub base_url: String,
    pub model_name: String,
    pub api_key: Option<String>,
    pub timeout_secs: u64,
}
FieldTypeDefaultDescription
base_urlString—Base URL including /v1 prefix (e.g., http://localhost:8000/v1)
model_nameString—Model identifier sent in the request payload
api_keyOption<String>NoneOptional bearer token
timeout_secsu64120HTTP request timeout in seconds

Note: The base_url should include the /v1 prefix if the server expects it. LocalLlmService appends /chat/completions to this URL.

GenerationConfig

Parameters that control LLM text generation.

pub struct GenerationConfig {
    pub temperature: f32,
    pub top_p: f32,
    pub max_tokens: u32,
    pub stop_sequences: Vec<String>,
    pub system_prompt: Option<String>,
}
FieldTypeDefaultDescription
temperaturef320.1Sampling temperature (0.0 = deterministic, 1.0+ = creative)
top_pf320.95Nucleus sampling threshold
max_tokensu322048Maximum tokens to generate
stop_sequencesVec<String>[]Stop sequences that halt generation
system_promptOption<String>NoneSystem message prepended to the conversation

Using Defaults

let config = GenerationConfig::default();
// temperature: 0.1, top_p: 0.95, max_tokens: 2048

Custom Configuration

let config = GenerationConfig {
    temperature: 0.3,
    top_p: 0.9,
    max_tokens: 4096,
    system_prompt: Some("You are a precise analyst.".into()),
    ..Default::default()
};

RagConfig

Configuration for the RAG workflow engine. Useful when loading configuration from files.

pub struct RagConfig {
    pub collection_name: String,
    pub top_k: usize,
    pub min_score: Option<f32>,
    pub generation: GenerationConfig,
}
FieldTypeDefaultDescription
collection_nameString—Vector store collection to search
top_kusize10Number of nearest-neighbor results
min_scoreOption<f32>NoneMinimum similarity score (0.0–1.0)
generationGenerationConfigDefaultLLM generation parameters

Loading from Files

All config structs support standard serde deserialization:

// From JSON file
let config: EmbeddingConfig = serde_json::from_str(&json_string)?;

// From TOML file
let config: LlmConfig = toml::from_str(&toml_string)?;

Loading from Environment Variables

A common pattern for production deployments:

let config = EmbeddingConfig {
    base_url: std::env::var("EMBEDDING_URL")
        .unwrap_or_else(|_| "http://localhost:8080".into()),
    model_name: std::env::var("EMBEDDING_MODEL")
        .unwrap_or_else(|_| "bge-large-en-v1.5".into()),
    batch_size: 32,
    timeout_secs: 30,
    api_key: std::env::var("EMBEDDING_API_KEY").ok(),
};