Timmy
46f961668f
Quality gates / quality (pull_request) Successful in 3m29s
A worker that answers /v1/models has not necessarily loaded its weights, so the app was offering an AI control that could only fail after the upload. Readiness now passes only once the pinned model is loaded, and the not-loaded state reaches manual Bristol entry in one tap. Verified against the real pinned llama.cpp build: it serves 503 while loading and the gate reports the distinct not-loaded state. Adds the reproducible container recipe (digest-pinned bases, pinned llama.cpp commit, hash-verified read-only weights, unprivileged, loopback-only publish), one shared post-load readiness gate used by the host bootstrap and the image healthcheck, and a target-GPU benchmark that refuses to write a receipt it did not measure. Refs #15
92 lines
4.6 KiB
JavaScript
92 lines
4.6 KiB
JavaScript
import { chromium } from 'playwright';
|
|
import assert from 'node:assert/strict';
|
|
import { mkdir } from 'node:fs/promises';
|
|
|
|
// Issue #15: readiness must pass only after the pinned model is loaded. A worker that answers
|
|
// /v1/models is reachable, but reachable is not loaded — the app must not offer an AI control
|
|
// that fails after the upload.
|
|
await mkdir('artifacts', { recursive: true });
|
|
const browser = await chromium.launch({ headless: true });
|
|
const context = await browser.newContext({ viewport: { width: 390, height: 844 }, deviceScaleFactor: 2, serviceWorkers: 'block' });
|
|
const page = await context.newPage();
|
|
const errors = [];
|
|
page.on('console', message => { if (message.type() === 'error') errors.push(message.text()); });
|
|
page.on('pageerror', error => errors.push(error.message));
|
|
let modelLoaded = false;
|
|
let analysisRequests = 0;
|
|
await page.route('**/api/vision-status', route => route.fulfill({
|
|
status: 200,
|
|
contentType: 'application/json',
|
|
body: JSON.stringify({
|
|
enabled: true,
|
|
profile: 'selfhost',
|
|
processor: 'self-hosted',
|
|
model: 'SmolVLM2-2.2B-Instruct',
|
|
providerReady: true,
|
|
modelSeen: modelLoaded,
|
|
}),
|
|
}));
|
|
await page.route('**/api/analyze', route => {
|
|
analysisRequests += 1;
|
|
return route.fulfill({
|
|
status: 200,
|
|
contentType: 'application/json',
|
|
body: JSON.stringify({
|
|
status: 'suggestion',
|
|
isStool: true,
|
|
bristolType: 4,
|
|
color: 'brown',
|
|
confidence: 0.83,
|
|
imageQuality: 'good',
|
|
observations: 'Smooth, formed appearance.',
|
|
warning: 'Visual suggestion only. Confirm it yourself; this is not a diagnosis.',
|
|
}),
|
|
});
|
|
});
|
|
|
|
await page.goto('http://127.0.0.1:4173', { waitUntil: 'networkidle' });
|
|
await page.evaluate(() => localStorage.clear());
|
|
await page.reload({ waitUntil: 'networkidle' });
|
|
|
|
// 1. Worker reachable, pinned model not loaded: no AI promise, no dead control, one action to manual.
|
|
await page.locator('[data-scan]').click();
|
|
await page.getByText(/model is not loaded/i).waitFor();
|
|
// Visual acceptance: the not-loaded state must be visually distinct from the ready and offline
|
|
// states, not an unstyled div that reads like either one of them.
|
|
const loadingStatus = page.locator('.model-status.loading');
|
|
assert.equal(await loadingStatus.count(), 1, 'the not-loaded state needs its own status treatment');
|
|
assert.equal(await loadingStatus.evaluate(el => getComputedStyle(el).backgroundColor), 'rgb(253, 243, 216)');
|
|
assert.equal(await page.getByText(/Self-hosted model ready/i).count(), 0, 'an unloaded model must never be reported ready');
|
|
assert.equal(await page.locator('#analyze-photo').count(), 0, 'no analyze control before the model is loaded');
|
|
await page.locator('#gallery-photo').setInputFiles('tests/fixtures/synthetic-type4.jpg');
|
|
await page.getByText(/Ready for your review/i).waitFor();
|
|
assert.equal(await page.getByText(/model is not loaded/i).count() > 0, true, 'the ready step still states why AI is unavailable');
|
|
assert.equal(await page.locator('.consent-card').count(), 0, 'no consent is requested while the model is not loaded');
|
|
assert.equal(await page.locator('#analyze-photo').count(), 0, 'no dead analyze control on the ready step');
|
|
const fallback = page.locator('#manual-from-scan');
|
|
assert.equal(await fallback.isVisible(), true);
|
|
assert.match(await fallback.innerText(), /without AI|manual/i);
|
|
await fallback.click();
|
|
assert.equal(await page.getByText('Pick the closest form').isVisible(), true, 'one action reaches manual Bristol entry');
|
|
assert.equal(analysisRequests, 0, 'nothing is uploaded before the pinned model is loaded');
|
|
await page.screenshot({ path: 'artifacts/worker-model-loading-mobile.png' });
|
|
await page.locator('#close-sheet').click();
|
|
|
|
// 2. The pinned model loads: the same flow now offers consent and analysis, and analysis works.
|
|
modelLoaded = true;
|
|
await page.locator('[data-scan]').click();
|
|
await page.getByText(/Self-hosted model ready/i).waitFor();
|
|
await page.locator('#gallery-photo').setInputFiles('tests/fixtures/synthetic-type4.jpg');
|
|
await page.getByText(/Ready for your review/i).waitFor();
|
|
assert.equal(await page.locator('.consent-card').isVisible(), true, 'consent returns once the model is loaded');
|
|
await page.locator('#ai-consent').check();
|
|
await page.locator('#analyze-photo').click();
|
|
await page.getByText(/83% confidence/i).waitFor();
|
|
assert.equal(analysisRequests, 1, 'analysis is offered exactly when the model is loaded');
|
|
await page.screenshot({ path: 'artifacts/worker-model-loaded-mobile.png' });
|
|
|
|
assert.deepEqual(errors, []);
|
|
await context.close();
|
|
await browser.close();
|
|
console.log('PASS readiness waits for the loaded pinned model: no AI control before load, full AI path after load');
|