Struct AiFocusTracker 

pub struct AiFocusTracker { /* private fields */ }

AI-driven focus tracker using LLM for intelligent analysis.

Implementations

impl AiFocusTracker

pub fn new(provider: Box<dyn Provider>, model: String) -> Self

Create a new AI-driven focus tracker.

Arguments

• provider - Provider for AI calls.
• model - Model name (recommend fast model like claude-haiku).

pub fn with_config( provider: Box<dyn Provider>, model: String, config: FocusTrackerConfig, ) -> Self

Create with custom configuration.

pub fn new_minimal(model: String) -> Self

Create a minimal tracker for background tasks.

pub fn current_focus(&self) -> Option<&ConversationFocus>

Get current focus.

pub fn set_focus(&mut self, focus: ConversationFocus)

Set current focus manually.

pub fn clear_focus(&mut self)

Clear current focus.

pub fn config(&self) -> &FocusTrackerConfig

Get configuration.

pub fn config_mut(&mut self) -> &mut FocusTrackerConfig

Get mutable configuration.

pub async fn analyze_message( &mut self, message: &Message, ) -> Result<FocusAnalysisResult>

Analyze a message’s relationship with current focus using AI.

This is the main method that replaces simple keyword matching with intelligent LLM-based analysis.

Arguments

• message - Message to analyze.

Returns

Analysis result with relevance score and focus update info.

pub async fn analyze_key_messages( &mut self, messages: &[Message], ) -> Result<Vec<(usize, FocusAnalysisResult)>>

Analyze a batch of key messages efficiently.

This method analyzes multiple messages but only calls AI for messages that are likely to affect focus (user messages, first/last messages).

Arguments

• messages - Messages to analyze.

Returns

List of analysis results for key messages.

pub fn detect_focus_fallback(&self, messages: &[Message]) -> ConversationFocus

Detect focus from messages (rule-based fallback).

This method provides a fallback when AI analysis is not available.

pub fn focus_score(&self, message: &Message) -> f32

Calculate focus score for a message using cached analysis.

Returns the relevance score from AI analysis if available, otherwise uses rule-based calculation.

pub fn create_focus_message(&self) -> Message

Create a focus message to inject into compressed conversation.