| description |
Index for MONITOR test-tier docs. One question per file. |
| tags |
|
| layer |
0 |
Three tiers, three files. L1 and L3 are opt-in; L2 is the always-on default.
| Tier |
What it exercises |
File |
| L1 — Playwright page smokes |
Every Play UI route mounts without throwing. Fast, no backend required. |
packages/ui/frontend/e2e/{pages,forge-redirects}.spec.ts |
| L2 — Full-stack replay suite |
The running backend over HTTP, seven surfaces (forge + copilot + long-form + …). |
REPLAYS.md |
| L3 — Live LLM roleplay |
Real browser + real backend + real Ollama driving a multi-turn session through the Play UI. Opt-in via E2E_ROLEPLAY=1. |
PLAYWRIGHT_LLM_ROLEPLAY.md |
| In-process full-loop |
Real character + sheet → bootstrap_story_scene → real SceneLoop, driven by InstructablePlayer. No HTTP. |
HARNESS_FULL_LOOP.md |
If a change touches the loop in-process, run the harness. If it touches
the live backend over HTTP, run the replays. If it touches the Play UI
rendering of the WS stream, run L3. The three answer different questions
— none subsumes another.