Struct GrokLlm 

pub struct GrokLlm { /* private fields */ }

XAI Grok LLM implementation.

Provides access to Grok models through the XAI API. The API is OpenAI-compatible with some differences in token counting.

Authentication

The API key can be provided explicitly or loaded from the XAI_API_KEY environment variable via from_env.

Model Selection

Common model names:

• grok-beta - Grok Beta
• grok-vision-beta - Grok Vision Beta

Token Counting Bug

Important: Grok has a known bug where reasoning_tokens in completion_tokens_details is NOT included in the top-level completion_tokens. This implementation automatically corrects this by adding reasoning_tokens to completion_tokens for accurate usage tracking.

Examples

use radkit::models::providers::GrokLlm;
use radkit::models::{BaseLlm, Thread};

// From environment variable
let llm = GrokLlm::from_env("grok-beta")?;

// With explicit API key
let llm = GrokLlm::new("grok-beta", "xai-...");

// Generate content
let thread = Thread::from_user("Explain quantum computing");
let response = llm.generate_content(thread, None).await?;
println!("{}", response.content().first_text().unwrap_or("No response"));

Implementations

impl GrokLlm

pub const API_KEY_ENV: &str = "XAI_API_KEY"

Environment variable name for the XAI API key.

pub fn new(model_name: impl Into<String>, api_key: impl Into<String>) -> Self

Creates a new Grok LLM instance with explicit API key.

Arguments

• model_name - The model to use (e.g., “grok-beta”)
• api_key - XAI API key

Examples

let llm = GrokLlm::new("grok-beta", "xai-...");

pub fn from_env(model_name: impl Into<String>) -> AgentResult<Self>

Creates a new Grok LLM instance loading API key from environment.

Reads the API key from the XAI_API_KEY environment variable.

Arguments

• model_name - The model to use (e.g., “grok-beta”)

Errors

Returns an error if the environment variable is not set or is empty.

Examples

let llm = GrokLlm::from_env("grok-beta")?;

pub fn with_base_url(self, base_url: impl Into<String>) -> Self

Sets a custom base URL for the API endpoint.

pub const fn with_max_tokens(self, max_tokens: u32) -> Self

Sets the maximum number of tokens to generate.

pub const fn with_temperature(self, temperature: f32) -> Self

Sets the temperature for generation (0.0 to 2.0).

Higher values produce more random outputs.

Trait Implementations

impl BaseLlm for GrokLlm

fn model_name(&self) -> &str

Returns the model identifier for this LLM instance.

fn generate_content<'life0, 'async_trait>( &'life0 self, thread: Thread, toolset: Option<Arc<dyn BaseToolset>>, ) -> Pin<Box<dyn Future<Output = AgentResult<LlmResponse>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait,

Generates content in response to a conversation thread.