v0.3.5 — Dashboard split, per-view endpoints, pipeline refactoring, static build (2026-04-20)

v0.3.5 — Dashboard split, per-view endpoints, pipeline refactoring, static build (2026-04-20)

Dashboard modularization

• Split dashboard.html (3023L monolithic) → css/tokens.css + 6 JS modules + analysis.html (895L)
• Dead code removed — _placeholder, _removed (~480 lines eliminated)
• Legend tab — inline content removed (598L), loads from /legend endpoint

Per-view endpoints

• Each view loads its own endpoint JSON (/mirror, /lens, /probe, /legend)
• Legacy analysis.json monolithic endpoint removed
• Cross-instance aggregation (all) added

Pipeline refactoring

• Reorganized filesystem/ → analysis/ with lib/, cli/, app/, data/, legend/
• Split analysis.py → scan.py, mirror.py, lens.py, probe.py
• Unified serve.py — GET /mirror, /lens, /probe, /legend + POST /refresh
• scan.py — session_dates, persona_roles, context_sizes in meta
• lens.py — totals, per-persona extras, silent weeks
• mirror.py — friction_records, map topology

Audit enhancement

• Dynamic artifact types from conventions.md
• PP1-PP4 persona checks (7 dimensions, workspace mapping)
• PA1-PA3 global artifact checks on all shared/
• IHM audit detail panel — clickable warn/fail entries
• Restructured audit IDs — PS/PF/AN/AR/AF/IS/IN/IR taxonomy

Tooling

• build_dist.py — static dashboard bundle generation, supports h2a-data structure
• audit.md reference + --protocol-only + --artifacts flags
• .gitignore — pycache, .pytest_cache, *.pyc

Decisions

• ADR — Remove monolithic analysis.json, per-view autonomous endpoints
• ADR — Pipeline split into specialized scripts (scan/mirror/lens/probe)