Enum DerivePolicy 

pub enum DerivePolicy {
    Legacy,
    Reset {
        threshold_tokens: usize,
    },
    Incremental {
        budget_tokens: usize,
    },
    OracleReplay {
        lookahead: usize,
    },
}

Per-session derivation strategy selector.

Defaults to DerivePolicy::Legacy so existing code paths that do not opt into a new policy get the Phase A behaviour unchanged.

Variants

Legacy

Phase A clone + experimental + enforce_on_messages + pairing. The historical pipeline.

Reset

Lu et al. (arXiv:2510.06727) reset-to-(prompt, summary).

When the request’s estimated token cost exceeds threshold_tokens, replace everything older than the last user turn with a single RLM-generated summary. The derived context for the next provider call contains [summary, active_task_tail...] so short continuation turns do not drop the task-defining prompt.

Fields

threshold_tokens: usize

Token budget that triggers the reset. Typically ~95 % of the model’s working context length.

Incremental

Liu et al. (arXiv:2512.22087) incremental select-then-pack derivation.

Score every entry against the latest user turn’s RelevanceMeta and greedy-pack the highest-scoring set into budget_tokens. The recent window is always retained; older entries are kept only when their relevance to the active task is high.

Fields

budget_tokens: usize

Per-call working-context budget. Older entries are dropped once the selected set would exceed this estimate.

OracleReplay

ClawVM replay oracle (Phase B step 22). Evaluation-only.

Given a recorded trace, picks representations with h-turn future-demand lookahead to minimise fault count. Not suitable for production — only for benchmarking online vs oracle gap.

Fields

lookahead: usize

Lookahead horizon in turns. ClawVM uses h=3.

Implementations

impl DerivePolicy

pub fn kind(&self) -> &'static str

Human-readable short name for logs, journal entries, and telemetry.

Examples

use codetether_agent::session::derive_policy::DerivePolicy;

assert_eq!(DerivePolicy::Legacy.kind(), "legacy");
assert_eq!(
    DerivePolicy::Reset { threshold_tokens: 0 }.kind(),
    "reset",
);

Trait Implementations

impl Clone for DerivePolicy

fn clone(&self) -> DerivePolicy

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for DerivePolicy

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

Formats the value using the given formatter.

impl Default for DerivePolicy

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for DerivePolicy

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for DerivePolicy

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

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl Copy for DerivePolicy