Struct DefaultNegotiator 

pub struct DefaultNegotiator { /* private fields */ }

Default negotiator implementation.

This implementation provides basic negotiation logic using an LLM:

• Uses structured LLM outputs to determine intent
• Asks clarifying questions when skill selection is ambiguous
• Transitions to task creation when a skill is clearly identified
• Rejects requests that are out of scope for available skills

Examples

use radkit::runtime::DefaultNegotiator;
use radkit::models::providers::AnthropicLlm;
use radkit::models::Content;

let llm = Arc::new(AnthropicLlm::from_env("claude-sonnet-4")?);

// Use default prompt
let negotiator = DefaultNegotiator::new(llm.clone());

// Override the prompt
let negotiator = DefaultNegotiator::new(llm)
    .with_prompt(Content::from_text("Custom negotiation prompt"));

Implementations

impl DefaultNegotiator

pub fn new(llm: Arc<dyn BaseLlm>) -> Self

Creates a new DefaultNegotiator with the given LLM provider.

Arguments

• llm - The LLM provider to use for negotiation decisions

pub fn with_prompt(self, prompt: Content) -> Self

Sets a custom negotiation prompt.

Arguments

• prompt - Custom system prompt to override the default

Trait Implementations

impl Clone for DefaultNegotiator

fn clone(&self) -> DefaultNegotiator

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Negotiator for DefaultNegotiator

fn negotiate<'life0, 'life1, 'life2, 'life3, 'async_trait>( &'life0 self, _auth_ctx: &'life1 AuthContext, agent_def: &'life2 AgentDefinition, _context_id: &'life3 str, content: Content, history: Vec<Message>, ) -> Pin<Box<dyn Future<Output = AgentResult<NegotiationDecision>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait, 'life3: 'async_trait,

Negotiates with the user to determine the next action.