LLM-Driven Extraction of Unstructured Data — Built for API Deployments & ETL Pipeline Workflows
Python ai-agentsdata-engineeringdocument-aigenerative-aiidpjson-extractionllmmcp-serverocrpdf-extractionprompt-engineeringstructured-output
import { defineConfig, devices } from "@playwright/test";
/**
* Playwright config for the `ui` rig group (tier: e2e).
*
* This is the tier the 348 Vitest tests structurally cannot cover: jsdom
* computes no layout, so overflow, positioning and visual state are
* unobservable there. Every UI defect found during the shadcn migration — a
* toggle rendering blank, a nav arrow anchored to the wrong month, a filter row
* breaking on a date range — was spotted by a human in a browser.
*
* Run it through the rig, which boots the platform and exports the URLs:
* python -m tests.rig run ui
*
* Or standalone against an already-running stack:
* UNSTRACT_FRONTEND_URL=http://localhost:3000 npx playwright test
*/
/*
* The rig exports UNSTRACT_FRONTEND_URL when it brings the platform up; the
* fallback matches the port compose publishes the frontend on.
*
* Deliberately no UNSTRACT_BACKEND_URL fallback. The frontend has its own
* origin (:3000 vs the backend's :8000), and the rig always exports the backend
* URL — so accepting it here would silently point every spec at the API and
* make the :3000 default unreachable in exactly the case it exists for.
*/
const baseURL = process.env.UNSTRACT_FRONTEND_URL ?? "http://localhost:3000";
export default defineConfig({
// Specs sit beside this config; `@playwright/test` is installed here too, so
// the suite is self-contained and resolves without reaching into frontend/.
testDir: ".",
/*
* `fullyParallel` off by default: these tests drive a shared platform with
* shared org state, so parallel specs can see each other's writes. Turn it on
* per-file once the suite has isolation.
*/
fullyParallel: false,
forbidOnly: Boolean(process.env.CI),
retries: process.env.CI ? 1 : 0,
workers: 1,
/*
* JUnit so the rig's `parse_junit` reads these results exactly like every
* other group's. The rig sets PLAYWRIGHT_JUNIT_OUTPUT_NAME to the group's
* junit.xml; the default keeps standalone runs from writing to the repo root.
*/
reporter: [
["list"],
[
"junit",
{
outputFile:
process.env.PLAYWRIGHT_JUNIT_OUTPUT_NAME ??
"reports/ui/junit.xml",
},
],
],
use: {
baseURL,
// Artefacts only for failures — a green suite should leave nothing behind.
trace: "on-first-retry",
screenshot: "only-on-failure",
video: "retain-on-failure",
},
projects: [
{
name: "chromium",
use: { ...devices["Desktop Chrome"] },
},
],
});