feat(delphi): replay harness H-A — schedule spec+slicer, Python driver, store, step comparer, CLI

[jucor]

commit-id:2324207e

---

Stack:

• ci: run postgres integration tests #2637
• feat(math): moderation + restart-seam replay coverage — mod_update parity, Q13-Q16 #2656
• docs(delphi): R1-parity goal docs, journal (2026-07-18 → 07-22 s3), quirks + divergence ledger #2626
• feat(delphi): Q12 carve-out — pin the cold-tick PCA start to ones in both replay drivers #2655
• feat(math): Q11 — replicate vectorz distance cancellation in legacy kmeans #2654
• feat(math): small-N degenerate parity — run the real math on tiny matrices in clojure-legacy mode #2653
• feat(delphi): Q10 carve-out — pin large-conv cutoffs to certify the deterministic full-PCA path #2652
• feat(delphi): replay loader resolves private datasets under real_data/.local #2651
• feat(math): Clojure hash-map iteration order for the in-conv greedy tie-break (legacy) #2650
• feat(math): clojure-legacy blob emission + selection parity — Clojure-exact math_main at step 0 #2649
• feat(delphi): acceptance canonicalization — suppress cross-engine ordering artifacts in certify comparer #2648
• feat(delphi): certify.py battery runner + focuser + fingerprint ledger (Spec A, Phase 0) #2647
• feat(delphi): prodclone extractor — feature-filtered dataset export (Spec B) #2646
• feat(math): un-mirror priority_metric (#2571 resolved) + Q2 prev-tick group-votes #2645
• feat(math): participant-ban leak replication in clojure-legacy mode (Q1) #2644
• feat(delphi): Clojure timing probe (Spec C) #2643
• feat(math): degenerate-tick group-clusterings overwrite + <2-participants edge (legacy mode) #2642
• refactor(math): move impl-flag resolver to polismath/utils/env_flags.py #2641
• feat(delphi): Python math poller service (phase 1) — Clojure math-container replacement #2625
• fix(delphi): Clojure-parity math writers + committing write path in PostgresClient #2624
• feat(math): in-conv greedy floor + persistent carry (PR-E) #2623
• feat(math): Clojure-exact base-cluster lineage + warm start (PR-C) #2622
• feat(math): replay harness H-B — Clojure Mode A driver (dev/replay.clj) + cross-language comparer shim #2621
• feat(math): group-k-smoother in clojure-legacy engine mode (PR-D smoother) #2620
• feat(math): PCA warm start in clojure-legacy engine mode (PR-B) #2640
• feat(math): add POLISMATH_ENGINE_MODE flag + prev-tick warm-start scaffolding (PR-A) #2618
• feat(delphi): replay harness H-A — schedule spec+slicer, Python driver, store, step comparer, CLI #2639 ⬅
• fix(delphi): cold-start generator copies FULL revote history (no DISTINCT ON dedup) #2638
• fix(math): correct comment-extremity sign in pca_project_cmnts (D1b) #2615
• docs(delphi): add math-algorithm history & routing-regression reference #2613

---

⚠️ Part of a stack created by spr. Do not merge manually using the UI - doing so may have unexpected results.

[Copilot]

Pull request overview

Adds a Delphi “replay harness” (Phase H-A) to deterministically replay exported vote streams on an explicit recompute schedule, persist per-step recordings under a local store, and compare recordings step-by-step using the existing regression comparer (with family-based reporting). This fits into the broader Delphi math-parity / schedule-inference work by providing a reproducible trace + comparison spine.

Changes:

• Introduce replay-harness core modules: dataset/types, schedule spec + slicer + presets, Python replay driver, recording store + provenance, and step comparer/reporting.
• Add a Click CLI (scripts/replay_driver.py) to run replays and compare recordings end-to-end.
• Add comprehensive unit + smoke tests covering schedule resolution/slicing, driver determinism and moderation seam guard, store layout/provenance, step comparison, and CLI behavior.

Reviewed changes

Copilot reviewed 13 out of 14 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
delphi/polismath/replay/init.py	Exposes replay-harness public types and documents phase scope.
delphi/polismath/replay/types.py	Adds core replay dataset/event types + stable time-sort and revote marking.
delphi/polismath/replay/schedule.py	Implements schedule spec round-trip, cut-slot resolution, slicing, moderation interleave, and presets.
delphi/polismath/replay/driver.py	Implements deterministic fold over scheduled vote batches, producing per-step records.
delphi/polismath/replay/real_data.py	Adds loader for exported real-data vote CSVs into a ReplayDataset.
delphi/polismath/replay/store.py	Implements on-disk recording layout, provenance capture, and numpy-aware JSON writing/loading.
delphi/polismath/replay/stepcompare.py	Wraps existing regression comparer to diff step blobs and format reports.
delphi/scripts/replay_driver.py	Adds CLI to run replays, store recordings, and compare two recordings.
delphi/tests/replay_harness/init.py	Introduces replay_harness test package.
delphi/tests/replay_harness/test_schedule.py	Tests cut resolution, slicing invariants, presets, moderation interleave, and JSON round-trip.
delphi/tests/replay_harness/test_driver.py	Tests end-to-end replay on vw, determinism guard, blob/extras completeness, and moderation-clear seam guard.
delphi/tests/replay_harness/test_store.py	Tests store layout, traversal rejection, provenance completeness, numpy JSON round-trip, and stale step cleanup.
delphi/tests/replay_harness/test_stepcompare.py	Tests divergence classification and directory-vs-directory compare behavior.
delphi/tests/replay_harness/test_cli.py	CLI smoke tests for run/compare and preset dataset single-load behavior.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.