Struct GenericCompletionModel 

pub struct GenericCompletionModel<Ext = OpenAICompletionsExt, H = Client> {
    pub model: String,
    /* private fields */
}

A chat-completions model over any OpenAICompatibleProvider extension. This is the advertised path for OpenAI-compatible providers; see the provider checklist in crate::providers.

Fields

model: String

Implementations

impl<Ext, H> GenericCompletionModel<Ext, H>

where Client<Ext, H>: HttpClientExt + Clone + 'static, Ext: Provider + OpenAICompatibleProvider + Clone + WasmCompatSend + 'static,

pub async fn raw_stream( &self, completion_request: CompletionRequest, ) -> Result<Pin<Box<dyn Stream<Item = Result<RawStreamingChoice<StreamingCompletionResponse<<Ext as OpenAICompatibleProvider>::StreamingUsage>>, CompletionError>> + Send>>, CompletionError>

Open a chat-completions stream whose terminal record stays provider-native.

This is the escape hatch for provider-specific terminal fields rig does not normalize. It shares the request builder, transport, telemetry, and error handling with CompletionModel::stream, which calls it and normalizes the terminal record — one network request either way.

impl<Ext, H> GenericCompletionModel<Ext, H>

where Client<Ext, H>: Debug + Clone + 'static, Ext: Provider + Clone + 'static,

pub fn new( client: Client<Ext, H>, model: impl Into<String>, ) -> GenericCompletionModel<Ext, H>

pub fn with_strict_tools(self) -> GenericCompletionModel<Ext, H>

Enable strict mode for tool schemas.

When enabled, tool schemas are automatically sanitized to meet OpenAI’s strict mode requirements:

• additionalProperties: false is added to all objects
• All properties are marked as required
• strict: true is set on each function definition

This allows OpenAI to guarantee that the model’s tool calls will match the schema exactly.

pub fn with_tool_result_array_content(self) -> GenericCompletionModel<Ext, H>

impl<Ext, H> GenericCompletionModel<Ext, H>

where Client<Ext, H>: HttpClientExt + Clone + WasmCompatSend + WasmCompatSync + 'static, Ext: Provider + OpenAICompatibleProvider + DebugExt + Clone + WasmCompatSend + WasmCompatSync + 'static, H: Clone + Default + Debug + WasmCompatSend + WasmCompatSync + 'static,

pub async fn raw_completion( &self, completion_request: CompletionRequest, ) -> Result<<Ext as OpenAICompatibleProvider>::Response, CompletionError>

Execute a chat completion and return the provider’s own wire response.

This is the escape hatch for provider-specific fields rig does not normalize. It shares the request builder, transport, telemetry, and error handling with CompletionModel::completion, which calls it and then applies the provider-local mapping — one network request either way.

The transport request id is not on the wire type and is dropped here; use Self::raw_completion_with_request_id when the typed route must reproduce everything completion returns.

pub async fn raw_completion_with_request_id( &self, completion_request: CompletionRequest, ) -> Result<(<Ext as OpenAICompatibleProvider>::Response, Option<String>), CompletionError>

Self::raw_completion plus the transport request id from the provider’s request-id response header (OpenAICompatibleProvider::REQUEST_ID_HEADER).

The pair exists because the wire type is substitutable — Ext::Response is whatever the compatible provider parses — so the transport id cannot live on it, while the normalized completion::CompletionResponse carries one. Without this method, raw_completion(..) followed by normalize would silently lack the provider_request_id that CompletionModel::completion reports — the typed escape hatch would not reproduce the normalized path. Reassemble with with_optional_provider_request_id.

impl<H> GenericCompletionModel<OpenRouterExt, H>

pub fn with_prompt_caching(self) -> GenericCompletionModel<OpenRouterExt, H>

Enable explicit prompt caching for supported OpenRouter models.

Adds cache_control: {"type": "ephemeral"} to the system-prompt block so subsequent turns that share the same system prefix can be billed at the cache-hit rate when the selected model/provider supports explicit cache breakpoints.

Trait Implementations

impl<Ext, H> Clone for GenericCompletionModel<Ext, H>

where Ext: Clone, H: Clone,

fn clone(&self) -> GenericCompletionModel<Ext, H>

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<Ext, H> CompletionModel for GenericCompletionModel<Ext, H>

where Client<Ext, H>: HttpClientExt + Clone + WasmCompatSend + WasmCompatSync + 'static, Ext: Provider + OpenAICompatibleProvider + DebugExt + Clone + WasmCompatSend + WasmCompatSync + 'static, H: Clone + Default + Debug + WasmCompatSend + WasmCompatSync + 'static,

fn capabilities(&self) -> ProviderCapabilities

Provider behavior a runtime should account for when preparing requests.

async fn completion( &self, completion_request: CompletionRequest, ) -> Result<CompletionResponse, CompletionError>

Generates a completion response for the given completion request.

async fn stream( &self, request: CompletionRequest, ) -> Result<StreamingCompletionResponse, CompletionError>

Streams a completion response for the given completion request.

fn completion_request( &self, prompt: impl Into<Message>, ) -> CompletionRequestBuilder<Self>

where Self: Sized + Clone,

Generates a completion request builder for the given prompt.

impl<Ext, H> ConstructCompletionModel<Client<Ext, H>> for GenericCompletionModel<Ext, H>

where Client<Ext, H>: Debug + Clone + 'static, Ext: Provider + Clone + 'static,

fn construct( client: &Client<Ext, H>, model: String, ) -> GenericCompletionModel<Ext, H>

Build this model from its provider client and a model identifier.