Enum ModelProvider 

pub enum ModelProvider {
    OpenAI {},
    OpenAICompatible {
        base_url: String,
        api_key: Option<String>,
        project_id: Option<String>,
    },
    AzureOpenAI {
        base_url: String,
        api_key: Option<String>,
        deployment: String,
        api_version: String,
    },
    Anthropic {
        base_url: Option<String>,
        api_key: Option<String>,
    },
    Gemini {
        base_url: String,
        api_key: Option<String>,
    },
    AzureAiFoundry {
        resource: String,
        api_key: Option<String>,
    },
    AwsBedrock {
        base_url: String,
        api_key: Option<String>,
    },
    GoogleVertex {
        base_url: String,
        api_key: Option<String>,
        project_id: Option<String>,
    },
    AlibabaCloud {
        base_url: String,
        api_key: Option<String>,
    },
    FalAi {
        api_key: Option<String>,
    },
}

Variants

OpenAI

OpenAICompatible

Fields

base_url: String

api_key: Option<String>

project_id: Option<String>

AzureOpenAI

Fields

base_url: String

api_key: Option<String>

deployment: String

api_version: String

Anthropic

Fields

base_url: Option<String>

api_key: Option<String>

Gemini

Fields

base_url: String

api_key: Option<String>

AzureAiFoundry

Fields

resource: String

Azure resource name (e.g. distri-tts-resource), not a URL. Every endpoint URL is derived from this — see ModelProvider::completion_url / ModelProvider::tts_url.

api_key: Option<String>

AwsBedrock

Fields

base_url: String

api_key: Option<String>

GoogleVertex

Fields

base_url: String

api_key: Option<String>

project_id: Option<String>

AlibabaCloud

Fields

base_url: String

api_key: Option<String>

FalAi

fal.ai — image-generation provider. The model id is the fal endpoint path (e.g. fal-ai/flux/dev); the gateway POSTs to https://fal.run/<model_id> with Authorization: Key <api_key>.

Fields

api_key: Option<String>

Implementations

impl ModelProvider

pub fn openai_base_url() -> String

pub fn anthropic_base_url() -> Option<String>

pub fn gemini_base_url() -> String

pub fn azure_api_version() -> String

pub fn alibaba_cloud_base_url() -> String

pub fn fal_ai_base_url() -> &'static str

fal.ai sync invocation root. The full URL is https://fal.run/<model_id>; auth is Authorization: Key <key>.

pub fn api_key_slot_mut(&mut self) -> Option<&mut Option<String>>

Mutable reference to this provider’s api_key slot, if any. Plain OpenAI returns None because it uses an env var directly.

pub fn base_url_slot_mut(&mut self) -> Option<&mut String>

Mutable reference to this provider’s base_url slot when it participates in endpoint-secret hydration. Anthropic’s Option<String> base_url is excluded — it has a default and no endpoint secret. Plain OpenAI has no base_url field.

pub fn provider_type(&self) -> ProviderType

Returns the provider type enum for this provider.

pub fn provider_id(&self) -> &str

Returns the provider ID string for secret lookup and “provider/model” format.

pub fn api_key_secret(&self) -> &'static str

The canonical secret-store key under which this provider’s API key lives.

Single source of truth. Every layer that needs to look up or validate the API key MUST go through this method — the gateway (ProviderClientConfig), workspace-level resolution (WorkspaceStore::resolve_model_settings), and the validator (required_secret_keys) all rely on it. The UI’s user-facing key list in default_models.json is kept in sync with this via a unit test.

pub fn endpoint_secret(&self) -> Option<&'static str>

The canonical secret-store key for this provider’s endpoint URL, or None if the provider has a fixed endpoint baked into the variant.

Only providers that require a tenant-specific endpoint (Azure, Bedrock, Vertex) return Some; everything else uses a default base URL.

pub fn azure_ai_foundry_base_url(resource: &str) -> String

OpenAI-compatible API base URL for an Azure AI Foundry resource name.

Azure AI Foundry exposes the OpenAI v1 API at https://<resource>.openai.azure.com/openai/v1 — one endpoint serving chat completions and OpenAI-style audio (TTS/STT). The Foundry project endpoint (*.services.ai.azure.com/api/projects/...) is the Agents SDK surface and is deliberately not used for model calls.

pub fn azure_ai_foundry_resource(&self) -> Option<&str>

Azure AI Foundry resource name, if this is that provider.

pub fn completion_url(&self) -> Option<String>

Resolved chat-completions base URL for providers whose endpoint is derived from config (Azure AI Foundry) rather than user-supplied. None means the caller should fall back to the provider’s own base_url/default.

pub fn tts_url(&self) -> Option<String>

Resolved TTS base URL. Azure AI Foundry serves TTS on the same OpenAI-compatible endpoint as chat today; kept as a separate method so a future non-OpenAI surface (e.g. Azure Speech) changes only here.

pub fn stt_url(&self) -> Option<String>

Resolved STT base URL — see ModelProvider::tts_url.

pub fn image_url(&self) -> Option<String>

Resolved image-generation base URL.

Azure AI Foundry exposes image generation on https://<resource>.services.ai.azure.com/openai/v1 — a different subdomain from chat/TTS (which use *.openai.azure.com). The image dispatcher consults this method so a Foundry resource routes to the right host without changing the chat path.

pub fn resolved_endpoint(&self) -> (Option<String>, Option<String>)

(base_url, api_key) for this provider — call after hydrate_creds. base_url is the OpenAI-compatible endpoint to probe; used by the /providers/test validation flow.

pub fn required_secret_keys(&self) -> Vec<&'static str>

Returns the required secret keys for this provider — i.e. keys that must resolve via the secret store or environment for the LLM call to succeed. If the provider has an inline api_key already configured on the provider variant, no secret lookup is required.

pub fn all_provider_definitions() -> Vec<ProviderSecretDefinition>

Returns all provider secret definitions — the built-in basics from default_models.json merged with any registered extensions.

pub fn well_known_models() -> Vec<ProviderModels>

Returns the well-known models grouped by provider — the built-in basics from default_models.json merged with registered extensions.

pub fn display_name(&self) -> &'static str

Get the human-readable name for a provider

pub fn otel_provider_name(&self) -> &'static str

OTel gen_ai.provider.name attribute value for this provider. Uses the semantic convention identifiers from the 2025 GenAI spec.

Trait Implementations

impl Clone for ModelProvider

fn clone(&self) -> ModelProvider

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ModelProvider

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

Formats the value using the given formatter.

impl Default for ModelProvider

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for ModelProvider

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl JsonSchema for ModelProvider

fn schema_name() -> Cow<'static, str>

The name of the generated JSON Schema.

fn schema_id() -> Cow<'static, str>

Returns a string that uniquely identifies the schema produced by this type.

fn json_schema(generator: &mut SchemaGenerator) -> Schema

Generates a JSON Schema for this type.

fn inline_schema() -> bool

Whether JSON Schemas generated for this type should be included directly in parent schemas, rather than being re-used where possible using the $ref keyword.

impl Serialize for ModelProvider

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

where __S: Serializer,

Serialize this value into the given Serde serializer.