Struct ProviderFactory 

pub struct ProviderFactory;

Provider factory for creating LLM and embedding providers.

Provides environment-based auto-detection and explicit provider selection.

Implementations

impl ProviderFactory

pub fn from_env() -> Result<(Arc<dyn LLMProvider>, Arc<dyn EmbeddingProvider>)>

Auto-detect and create providers from environment.

Priority

1. EDGEQUAKE_LLM_PROVIDER environment variable (explicit selection)
2. Auto-detect: OLLAMA_HOST → LMSTUDIO_HOST → OPENAI_API_KEY → Mock

Returns

Returns a tuple of (LLMProvider, EmbeddingProvider). In most cases, the same provider implementation is used for both.

Errors

Returns error if required configuration for selected provider is missing.

Examples

std::env::set_var("OLLAMA_HOST", "http://localhost:11434");
let (llm, embedding) = ProviderFactory::from_env()?;
assert_eq!(llm.name(), "ollama");

pub fn create( provider_type: ProviderType, ) -> Result<(Arc<dyn LLMProvider>, Arc<dyn EmbeddingProvider>)>

Create specific provider type.

Arguments

• provider_type - The type of provider to create

Returns

Returns a tuple of (LLMProvider, EmbeddingProvider).

Errors

Returns error if required configuration for the provider is missing.

pub fn create_with_model( provider_type: ProviderType, model: Option<&str>, ) -> Result<(Arc<dyn LLMProvider>, Arc<dyn EmbeddingProvider>)>

OODA-04: Create specific provider type with a model override.

Like create() but allows specifying a model name instead of using defaults. Useful for CLI where user specifies --provider openrouter --model mistral/model.

Arguments

• provider_type - The type of provider to create
• model - Optional model name to use instead of provider default

Returns

Returns a tuple of (LLMProvider, EmbeddingProvider).

Errors

Returns error if required configuration for the provider is missing.

pub fn from_config( config: &ProviderConfig, ) -> Result<(Arc<dyn LLMProvider>, Arc<dyn EmbeddingProvider>)>

OODA-200: Create provider from TOML configuration.

This is the primary entry point for custom providers defined in models.toml. Supports both native providers (OpenAI, Ollama) and OpenAI-compatible APIs.

Arguments

• config - Provider configuration from models.toml

Returns

Returns a tuple of (LLMProvider, EmbeddingProvider).

Errors

Returns error if:

• Required API key environment variable is not set
• Base URL is not configured for OpenAI-compatible providers

Examples

use edgequake_llm::{ModelsConfig, ProviderFactory};

let config = ModelsConfig::load()?;
let provider_config = config.get_provider("zai").unwrap();
let (llm, embedding) = ProviderFactory::from_config(provider_config)?;

pub fn from_config_with_model( config: &ProviderConfig, model_name: Option<&str>, ) -> Result<(Arc<dyn LLMProvider>, Arc<dyn EmbeddingProvider>)>

Create provider from configuration with a specific model override.

This is used when the user selects a specific model via /model command that differs from the provider’s default model.

Arguments

• config - Provider configuration from models.toml
• model_name - Optional model name to use instead of the default

Returns

Tuple of (LLM provider, Embedding provider)

pub fn create_azure_openai() -> Result<(Arc<dyn LLMProvider>, Arc<dyn EmbeddingProvider>)>

Create Azure OpenAI provider from environment variables.

Tries AZURE_OPENAI_CONTENTGEN_* first (common enterprise naming), then falls back to AZURE_OPENAI_* standard variables.

Environment variables (CONTENTGEN variant):

• AZURE_OPENAI_CONTENTGEN_API_ENDPOINT → endpoint URL
• AZURE_OPENAI_CONTENTGEN_API_KEY → API key
• AZURE_OPENAI_CONTENTGEN_MODEL_DEPLOYMENT → deployment name
• AZURE_OPENAI_CONTENTGEN_API_VERSION → (optional)

Environment variables (standard variant):

• AZURE_OPENAI_ENDPOINT → endpoint URL
• AZURE_OPENAI_API_KEY → API key
• AZURE_OPENAI_DEPLOYMENT_NAME → deployment name
• AZURE_OPENAI_EMBEDDING_DEPLOYMENT_NAME → (optional)
• AZURE_OPENAI_API_VERSION → (optional)

pub fn embedding_dimension() -> Result<usize>

Get embedding dimension for current provider configuration.

Useful for configuring vector storage with the correct dimension.

Examples

std::env::set_var("EDGEQUAKE_LLM_PROVIDER", "ollama");
let dim = ProviderFactory::embedding_dimension()?;
assert_eq!(dim, 768); // embeddinggemma dimension

pub fn create_embedding_provider( provider_name: &str, model: &str, _dimension: usize, ) -> Result<Arc<dyn EmbeddingProvider>>

Create an embedding provider from workspace configuration.

This is used to create workspace-specific embedding providers for query execution. The provider is configured with the workspace’s embedding model and dimension.

@implements SPEC-032: Workspace-specific embedding in query process

Arguments

• provider_name - Provider type (e.g., “openai”, “ollama”, “lmstudio”, “vscode-copilot”, “mock”)
• model - Embedding model name (e.g., “text-embedding-3-small”, “embeddinggemma:latest”)
• dimension - Embedding dimension (e.g., 1536, 768)

Returns

Returns an Arc<dyn EmbeddingProvider> configured for the workspace.

Errors

Returns error if the provider type is unknown or required configuration is missing.

Examples

let provider = ProviderFactory::create_embedding_provider(
    "ollama",
    "embeddinggemma:latest",
    768,
)?;
assert_eq!(provider.dimension(), 768);

pub fn create_llm_provider( provider_name: &str, model: &str, ) -> Result<Arc<dyn LLMProvider>>

Create an LLM provider from workspace configuration.

This is used to create workspace-specific LLM providers for ingestion/extraction. The provider is configured with the workspace’s LLM model.

@implements SPEC-032: Workspace-specific LLM in ingestion process

Arguments

• provider_name - Provider type (e.g., “openai”, “ollama”, “lmstudio”, “mock”)
• model - LLM model name (e.g., “gpt-4o-mini”, “gemma3:12b”)

Returns

Returns an Arc<dyn LLMProvider> configured for the workspace.

Errors

Returns error if the provider type is unknown or required configuration is missing.

Examples

let provider = ProviderFactory::create_llm_provider(
    "ollama",
    "gemma3:12b",
)?;
assert_eq!(provider.model(), "gemma3:12b");