Milestone 4.5.D.3 — Benchmark Sign-off

Status: Planned
Goal: Track D — Exit Gate
Phase: 4.5 — Intelligence Layer
Estimated effort: 1 day
Track: Track D — Exit Gate

Why This Milestone Exists

Extends the existing aggregate_metrics.py → regression_gate.py → build_benchmark_data.py pipeline with Phase 4.5-specific targets, ensuring intelligence features don't degrade proxy performance.

Benchmark Targets

Benchmark	Target	Rationale
Fingerprint computation (100 sessions)	p95 < 500ms	Async worker budget
Drift comparison (fingerprint vs. baseline)	p95 < 50ms	Per-fingerprint-cycle, not per-request
Regression suite (50 scenarios, mocked judge)	< 30s wall clock	Operator-facing wait in promotion UI
Proxy overhead with Phase 4.5 enabled	<20ms p99 non-provider	Standing regression gate since Phase 2

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):

Evaluation / benchmarks
Fingerprinting / drift
Directive regression / promotion

Success signals

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

All 4 benchmarks passing in CI
Proxy overhead re-certified at <20ms p99 with fingerprinting/drift enabled
Results committed to benchmark data pipeline

Prerequisites

All Track A–C milestones complete
Existing benchmark harness operational