Proposed specificationSource-located exemplarH3

Lineage and threshold compaction

Bound history while retaining useful state.

JOIN POINTcontext_projectionRECORD STATUSCandidate · v0.1.0

UML class view

The proposed contract drawn as participant roles, received calls, and numbered dependencies.

Object interactions

The same roles and calls as a class collaboration and a numbered message sequence.

Object interaction diagram for Lineage and threshold compaction, showing role collaboration and a numbered message sequence — FIG. P27 / OBJECT INTERACTIONS
Open full SVG Download PDF

Pointcut

Where the mechanism applies. All four dimensions belong to the proposed match, not just its vocabulary.

Context: History approaches a context or count threshold before a model call.
Trigger: A compaction policy activates.
Structure: Threshold, selection strategy, summarizer or truncator, and lineage metadata.
Effect: Model input is reduced while retained and removed evidence remains attributable.

Protected contract

The obligation that clients, operators, and auditors should be able to rely on.

Requires: Context usage, protected items, and eviction policy are available.
Ensures: The replacement fits the target and records what replaced which history.
Invariant: Required system, safety, or current-task evidence is not silently evicted.
Violation: Summary failure, threshold conflict, or empty projection falls back explicitly.

Negative control

A generic text truncate utility is not compaction without lineage.

Variation & lifecycle

The hot spot is admitted only where the protected semantics survive.

Permitted variation

Message count, token budget, oldest-first, protected groups, remote or local.

Test obligation

Test threshold edge, protected item, failure, lineage, and post-compaction facts.

Evolution question

Did summarization become thresholded and lineage-aware?

Prompt + code design

An illustrative prompt and a small design sketch for the proposed contract. Names and APIs are invented; this is not a repository excerpt.

PROMPT SKETCH

Compact context against a target budget while protecting declared items. Record which history was replaced by which summary, and keep the lineage queryable after compaction.

CODE DESIGN / PYTHON

def compact(history, policy, target_tokens):
    protected, evictable = history.split(policy.protected_ids)
    summary = summarize(evictable, max_tokens=target_tokens - protected.tokens)
    replacement = ContextSummary(
        text=summary,
        replaces=[item.id for item in evictable],
        protected=[item.id for item in protected],
    )
    return history.replace(evictable, replacement)

Illustrative design sketch. Compaction is a provenance operation, not just a shorter string. This example is a teaching aid, not implementation evidence or a conformance claim.

Detection & matching

The frozen rule that retrieves this record, the canonical source exemplar, and the control that should stay unmatched.

Join point: context_projection
Source locator: Barbaste et al., §9.5, p. 32 (“Threshold Compaction”); §9.5, p. 33 (Hermes, “Lineage Compaction”); merged label
Lexical cue: (?i)\b(?:compact(?:ion|ed|or)?|summari[sz](?:e|ation)|truncate[_-]?history|token[_-]?limited[_-]?history)\bFrozen expression, v0.1.0. A match is a retrieval cue, not a confirmed implementation.
Aliases: None
Canonical exemplar: anomalyco/opencodepackages/core/src/session/compaction.ts:298Support: packages/core/test/session-native-compaction.test.ts:213
Negative control: A generic text truncate utility is not compaction without lineage.

Scope-anchored review. pohe-agent analyze <repo> --pattern P27 anchors the cue to its enclosing declaration and checks a separate operational signature for all four pointcut dimensions. Missing roles and negative-control risks remain visible; a structural match is not E2/E3. Calibration report ↗

Repository files

Up to five public files retrieved by the frozen cue and pinned to the audited commit. A match is a place to inspect, not a confirmed implementation.

01anomalyco/opencodepackages/core/src/session/compaction.ts:298CANONICAL SOURCE 02aaif-goose/goosecrates/goose/examples/agent.rs:65RETRIEVAL CUE 03affaan-m/ECCskills/agent-self-evaluation/scripts/evaluate.py:267RETRIEVAL CUE 04agno-agi/agnolibs/agno/agno/db/base.py:2284RETRIEVAL CUE 05Aider-AI/aideraider/args.py:189RETRIEVAL CUE

Evidence & provenance

Each layer answers a different question. Do not substitute retrieval for review.

Source-located exemplar

A concrete source anchor is available for this proposed record. It is one exemplar, not evidence that all repositories—or every contract field—conform.

Review boundary: this candidate and its specification come from an AI-assisted study. Independent human validation remains open.

ANY-EVIDENCE FAMILY COVERAGE35/36

97.2% of repository families in the frozen panel have at least one current cue, historical cue, or canonical source layer. This is not confirmed prevalence.

Current lexical-cue repositories: 39 / 41

Canonical source anchoranomalyco/opencode 885df74024b6packages/core/src/session/compaction.ts:298

Supporting surfacepackages/core/test/session-native-compaction.test.ts:213

Threshold, background, and native compaction.

Exploratory retrieval cue Regex / v0.1.0

This expression is a discovery instrument only. A match is a retrieval cue—not a confirmed implementation, pointcut match, or contract proof.