Struct OpenRouter 

pub struct OpenRouter { /* private fields */ }

Implementations

impl OpenRouter

pub fn from_key_auto(key: String) -> Self

pub fn openrouter_flavor(key: String) -> Self

pub fn openai(key: String) -> Self

pub fn opencode_go(key: String) -> Self

pub fn openai_codex(key: String) -> Self

pub const fn backend_tag(&self) -> BackendTag

pub fn is_openrouter(&self) -> bool

Video gen is OpenRouter-only (not OpenAI/Codex/OpenCode).

pub async fn generate_video(&self, req: VideoRequest) -> Result<(Vec<u8>, f64)>

Generate a video via OpenRouter’s /api/v1/videos endpoint. Submits the job, polls every 10s (up to 6 min), downloads on completion. Returns (mp4_bytes, cost_in_usd).

pub const fn default_utility_model(&self) -> &'static str

pub const fn default_research_model(&self) -> &'static str

pub const fn default_embedding_model(&self) -> &'static str

pub const fn default_video_gen_model(&self) -> &'static str

pub const fn default_image_gen_model(&self) -> &'static str

pub async fn list_models(&self) -> Result<Vec<Model>>

Fetch the live model catalog. No hardcoded list except Codex, whose subscription endpoint does not expose the normal OpenAI models catalog.

pub async fn complete( &self, model: &str, messages: Vec<ChatMessage>, ) -> Result<String>

One-shot, non-streaming completion. Used for short utility calls like generating a session topic/slug. Returns the assistant’s message text.

pub async fn complete_with_usage( &self, model: &str, messages: Vec<ChatMessage>, ) -> Result<Completion>

One-shot completion with the provider’s usage object when available. Keeping usage attached to the text lets background workflow stages use the same cache accounting as the interactive stream.

pub async fn complete_with_params( &self, model: &str, messages: Vec<ChatMessage>, params: &ChatParams, ) -> Result<Completion>

One-shot completion with request parameters and usage accounting.

pub async fn describe_image( &self, model: &str, image_data_url: &str, ) -> Result<String>

One-shot, non-streaming vision call: describe image_data_url with model.

pub async fn generate_image( &self, model: &str, prompt: &str, size: &str, image_data: Option<&[u8]>, ) -> Result<(Vec<u8>, String)>

Generate an image via OpenRouter’s dedicated /api/v1/images endpoint. Returns (decoded_bytes, file_extension) — the caller should use the returned extension (e.g. "png", "jpg", "webp") for the saved file.

pub async fn ocr_page( &self, model: &str, image_data_url: &str, ) -> Result<String>

One-shot, non-streaming vision call: transcribe a scanned page image.

pub async fn embed( &self, model: &str, inputs: Vec<String>, ) -> Result<Vec<Vec<f32>>>

Embed inputs with model (OpenAI-format /embeddings endpoint); returns one vector per input, in order.

pub fn stream_chat( &self, model: String, messages: Vec<ChatMessage>, params: ChatParams, tools: Vec<ToolDef>, toolbox: Arc<dyn ToolExecutor>, max_tool_iters: usize, ) -> (UnboundedReceiver<StreamEvent>, AbortHandle)

Start a streaming completion. Spawns a task that pushes tokens over the returned channel; the UI loop drains it alongside keypresses. If the model calls a tool, the task runs it via toolbox and continues the conversation, bounded by max_tool_iters round-trips.

Trait Implementations

impl Clone for OpenRouter

fn clone(&self) -> OpenRouter

Returns a duplicate of the value.

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

Performs copy-assignment from source.