Milestone 5.B.2 — Graph-Boosted Scoring

Status: Planned
Goal: Track B — Graph-Aware Retrieval
Phase: 5 — Advanced Retrieval
Estimated effort: 2 days
Track: Track B — Graph-Aware Retrieval

Why This Milestone Exists

Two concrete signals from lineage: (1) superseded memories should never outrank their successors, and (2) unresolved contradictions (both pending_review) should not both be packed into context — that actively confuses the model. The packer needs a contradiction-edge check before formatting.

Orientation (indicative)

Named paths, package layouts, libraries, schemas, env vars, and commands anywhere on this page are rough sketches for orientation — inspiration and a baseline, not a required change list.

During implementation, expect to:

open the live tree and follow existing patterns before inventing new ones
research current constraints (latency, tenancy, deploy shape, libraries) more deeply than this page can
advance the design beyond the sketch where measurement or code reality says so
land work in different filenames, merged packages, deferred docs, or new surfaces when the situation calls for it

Prefer outcomes over matching any particular file tree or command sequence.

Areas that may be involved (situational — not a checklist):

Graph / lineage traversal
Memory service / repositories
MCP server surface

Success signals

Outcome-oriented signals that the milestone is in good shape. Exact filenames, package layouts, and commands may differ from any sketches above.

Contradiction filter: two candidates with contradicts edge and no resolution → only higher-scored survives
Golden-output regression test extended with contradiction-pair fixture
ADR-0049 written: "Why we filter unresolved contradictions at pack time"

Prerequisites

Milestone 5.B.1 (Graph traversal query layer)
Context assembly packer (Phase 3.5)