Struct MetaAdapter 

pub struct MetaAdapter {
    pub top_k: usize,
    /* private fields */
}

Prototype store for task-level meta-adaptation.

Examples

use std::collections::HashMap;
use lmm_agent::cognition::learning::meta::MetaAdapter;
use lmm_agent::cognition::learning::q_table::ActionKey;

let mut a = MetaAdapter::new(5);
let offsets = HashMap::from([(ActionKey::Expand, 0.7)]);
a.record_episode("rust async await", offsets, 1.0);
let r = a.adapt("async rust futures");
assert!(r.contains_key(&ActionKey::Expand));

Fields

top_k: usize

Maximum number of prototypes considered per adapt call.

Implementations

impl MetaAdapter

pub fn new(top_k: usize) -> Self

Constructs a new MetaAdapter with the given top-k lookup limit.

pub fn len(&self) -> usize

Returns the number of stored prototypes.

pub fn is_empty(&self) -> bool

Returns true when no prototypes have been stored.

pub fn record_episode( &mut self, goal: &str, offsets: HashMap<ActionKey, f64>, avg_reward: f64, )

Stores a new episode as a task prototype.

goal is tokenised into a HashSet<String>. If a nearly identical prototype already exists (Jaccard ≥ 0.9), its offsets are blended rather than creating a duplicate.

Arguments

• goal - Natural-language task goal.
• offsets - Per-action Q-value offsets from the completed episode.
• avg_reward - Mean reward across the episode steps.

pub fn adapt(&mut self, goal: &str) -> HashMap<ActionKey, f64>

Returns weighted Q-offset priors for a new goal.

Finds the top-K most similar prototypes, weights their offsets by similarity × avg_reward, and returns the normalised blend.

Returns an empty map when no prototypes exist.

Examples

use std::collections::HashMap;
use lmm_agent::cognition::learning::meta::MetaAdapter;
use lmm_agent::cognition::learning::q_table::ActionKey;

let mut a = MetaAdapter::new(3);
a.record_episode("machine learning", HashMap::from([(ActionKey::Expand, 0.6)]), 0.7);
let out = a.adapt("deep learning models");
assert!(!out.is_empty());

pub fn prototypes(&self) -> &[TaskPrototype]

Returns a slice over all stored prototypes.

Trait Implementations

impl Clone for MetaAdapter

fn clone(&self) -> MetaAdapter

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for MetaAdapter

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for MetaAdapter

fn default() -> MetaAdapter

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for MetaAdapter

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for MetaAdapter

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.