playwright

name

playwright

description

Run Playwright browser E2E against a REAL portal, not the dev Monolith. The Monolith has DevLogin but no language model; the running `memex` stack has a model but uses Entra OAuth and is the user's data. This skill stands up a throwaway, DevLogin portal on Colima k3s — built from the CURRENT working tree, with its OWN database (memex_e2e), reusing the existing memex Postgres + host Ollama + ingress/TLS — then points Playwright at it over the ingress (reverse proxy). Use when an E2E must actually log in, render Blazor/SignalR, AND execute a chat round (or any flow needing a model); or when a test "skips execution" for lack of a model. One command: `memex-local e2e up` → `e2e test` → `e2e down`. Built on memex-local / LocalColimaMac.

/playwright — deploy a throwaway portal on Colima, then drive it with Playwright

Never run a chat/model E2E against the dev Monolith — it has DevLogin but no language model, so Send is gated and the test can only assert the composer renders, never that it executes. And never point Playwright at the user's memex portal — it authenticates through Entra (no /dev/signin) and holds their real data.

Instead: build the image from the working tree, deploy a dedicated DevLogin portal on Colima k3s, and drive THAT. It reuses the running memex stack's Postgres, host Ollama and ingress/TLS, but runs against its own database (memex_e2e) with DevLogin on, behind the ingress (reverse proxy) at https://e2e.memex.localhost:8444. It is additive — it never touches the memex release, DB, or config.

Canonical references:

• LocalColimaMac.md — the Colima k3s stack this builds on (§15 = the e2e portal).
• deploy/homebrew/bin/memex-local — the CLI; the e2e subcommand is the whole flow.
• test/MeshWeaver.Portal.E2E.Test/PortalFixture.cs — the DevLogin fixture every E2E uses.

The flow (one CLI, three steps)

# 0. Prereq: the memex stack must already be up (provides PG, Ollama, ingress, TLS):
memex-local up                      # or: memex-local status   (skip if already running)

# 1. Build the working tree → image, create memex_e2e, migrate, deploy, reverse-proxy:
memex-local e2e up                  # add --skip-build to reuse the last image (fast iterate)

# 2. Run a Playwright test against the e2e portal (E2E_BASE_URL + DevLogin are preset):
memex-local e2e test HomeChatExecuteTest    # default filter is HomeChatExecuteTest

# 3. Tear down — delete the e2e objects + drop the e2e DB (keeps the memex stack):
memex-local e2e down                # add --keep-db to keep memex_e2e between runs

Run the CLI from the repo (it auto-resolves the chart/repo), or MEMEX_REPO=$PWD memex-local … when invoking deploy/homebrew/bin/memex-local directly. memex-local e2e status / e2e logs inspect the running instance.

What e2e up does (and why each step)

1. image_build_local — dotnet publish -t:PublishContainer for the portal + migration (native arm64, into Colima's image store). --skip-build reuses the last build for fast loops.
2. Create memex_e2e in the existing memex-postgres (a separate database — never memex).
3. Migrate it (a one-off Job on the migration image). The portal's DbVersionGate refuses to start against an un-migrated DB, so this must finish first.
4. Deploy memex-e2e-portal (Deployment + Service + Ingress) in the memex namespace, reusing memex-portal-config / memex-portal-secrets via envFrom, overriding only: ConnectionStrings__memex → memex_e2e, and Authentication__EnableDevLogin=true. Clustering is Localhost (its own in-process silo) so it never joins memex's cluster.
5. Reverse proxy — an Ingress for e2e.memex.localhost (covered by the *.memex.localhost mkcert cert), plus a port-forward of ingress-nginx :443 to host :8444.

Driving it with Playwright

PortalFixture authenticates via DevLogin (POST /dev/signin?personId=Roland) and sets IgnoreHTTPSErrors=true, so the self-signed *.memex.localhost cert is fine. memex-local e2e test exports E2E_BASE_URL=https://e2e.memex.localhost:8444 and E2E_USER=Roland for you. To run a test by hand:

E2E_BASE_URL=https://e2e.memex.localhost:8444 E2E_USER=Roland \
  dotnet test test/MeshWeaver.Portal.E2E.Test --filter "FullyQualifiedName~<TestName>"

When writing a new browser test, follow the existing ones:

• Put it in test/MeshWeaver.Portal.E2E.Test, [Collection("portal-e2e")], ctor takes PortalFixture.
• First line: Assert.SkipUnless(fixture.Available, fixture.SkipReason) — so it no-ops when E2E is off.
• await fixture.NewAuthenticatedContextAsync() for a logged-in browser context; seed mesh content it can't create via the UI with MintTokenAsync + CreateNodeAsync.
• Assert on rendered DOM, not typed-but-unsent input (the HomeChatExecuteTest lesson: matching the editor's own text false-passes). Wait for a real message bubble / state change.
• A model-dependent step should detect "No language model is available" and Assert.Skip rather than false-pass — but under THIS skill a model IS present, so execution actually runs.

Iterate fast

After a code change: memex-local e2e up rebuilds + rolls the portal (the DB/migration are idempotent). For a UI-only change with the image unchanged, --skip-build skips the long publish.

Troubleshooting

• namespace 'memex' not found → run memex-local up first; e2e reuses that stack.
• Portal not ready / e2e logs shows a DB error → the migration Job didn't complete; check kubectl -n memex logs job/memex-e2e-migration.
• Send still gated in the test → host Ollama isn't serving the model. ollama list must show the chat model (qwen3.6-code); the portal addresses it as http://ollama:11434/v1.
• 502 / not 200 at the URL → the ingress port-forward dropped; memex-local e2e status restarts it, or kubectl -n ingress-nginx port-forward svc/ingress-nginx-controller 8444:443.
• Leftover env → memex-local e2e down (idempotent). It only removes e2e objects + the e2e DB.