robbond
79ea2f6824
Add explicit reasoning guidance for normalising counts by exposure/denominator, distinguishing total count from rate, and avoiding correlation-as-causation errors. Changes: - prompts/reconstruct-v0.3.md: new prompt with normalisation discipline - lib/reconstruction/prompt.js: v0.3 loader + env var override support - lib/analysis.js: defer DEFAULT_PROMPT_VERSION to prompt module (defaults to v0.3) - PROMPT_VERSIONS extended to [v0.1, v0.2, v0.3] - tests/v03-reasoning.test.js: 34 focused tests covering prompt loading, schema validation, guidance completeness, and target scenario fixture - playwright.config.js + tests/smoke.test.js: minimal UI smoke test for browser rendering - package.json: add @playwright/test as devDependency Default switches to v0.3; v0.2 selectable via promptVersion or RECONSTRUCTION_PROMPT_VERSION env var.
56 lines
2.4 KiB
JavaScript
56 lines
2.4 KiB
JavaScript
import { test, expect } from "@playwright/test";
|
|
|
|
test("v0.3 UI smoke test with live model response", async ({ page }) => {
|
|
await page.goto("http://localhost:3000");
|
|
|
|
// Page should load without error
|
|
await expect(page.getByText(/Confidence Engine/i)).toBeVisible();
|
|
|
|
// Type the scenario
|
|
const textarea = page.locator("textarea[placeholder*='Describe']");
|
|
await textarea.fill("Complaints increased by 35% while production increased by 40%.");
|
|
|
|
// Button should be enabled
|
|
await expect(page.getByRole("button", { name: /Analyse/i })).toBeEnabled();
|
|
|
|
// Click Analyse and wait for diagnostics panel
|
|
await page.getByRole("button", { name: /Analyse/i }).click();
|
|
|
|
// Wait for result section (ReconstructionView rendered)
|
|
await expect(page.getByRole("heading", { name: /Next Question/i })).toBeVisible({ timeout: 180000 });
|
|
|
|
// Take screenshot of result page
|
|
await page.screenshot({ path: "tests-results/smoke-v0.3.png", fullPage: true });
|
|
|
|
// Verify diagnostics panel exists and contains relevant info
|
|
const diagPanel = page.locator('details summary').first();
|
|
if (await diagPanel.isVisible()) {
|
|
console.log("Raw response viewer:", await diagPanel.innerText().catch(() => "not visible"));
|
|
}
|
|
|
|
// Get full body text for verification
|
|
const bodyText = await page.locator("body").innerText();
|
|
|
|
console.log("\n=== UI Smoke Test Results ===");
|
|
console.log("Page title:", await page.title());
|
|
console.log("Body content length:", bodyText.length);
|
|
|
|
// Check key content indicators
|
|
const hasNextQ = bodyText.includes("Next Question");
|
|
const hasComplaints = bodyText.includes("Complaint") || bodyText.includes("complaint");
|
|
const hasProduction = bodyText.includes("production") || bodyText.includes("Production");
|
|
const hasRateContext = bodyText.toLowerCase().includes("rate") ||
|
|
bodyText.toLowerCase().includes("unit") ||
|
|
bodyText.toLowerCase().includes("denominator") ||
|
|
bodyText.toLowerCase().includes("per-unit");
|
|
|
|
console.log("Has Next Question heading:", hasNextQ);
|
|
console.log("Has complaints reference:", hasComplaints);
|
|
console.log("Has production reference:", hasProduction);
|
|
console.log("Has rate context (rate/unit/denominator):", hasRateContext);
|
|
|
|
// Basic structural checks
|
|
expect(bodyText.length).toBeGreaterThan(200);
|
|
expect(hasNextQ).toBe(true);
|
|
}, { timeout: 300000 });
|