v4.0.0

.claude/skills/generate-e2e-tests/SKILL.md

@@ -0,0 +1,335 @@
+---
+name: generate-e2e-tests
+description: Generate comprehensive Playwright e2e tests from a .netcanvas protocol file and an optional recording. Invoke with /generate-e2e-tests <protocol-path> [recording-path]
+user-invocable: true
+---
+
+# Generate E2E Tests from Protocol
+
+You are generating comprehensive Playwright e2e tests for a Fresco interview protocol.
+
+## Inputs
+
+- **Protocol path**: `$1` — path to a `.netcanvas` file (ZIP containing `protocol.json`)
+- **Recording path** (optional): `$2` — path to a recording directory (contains `actions.jsonl`, `SESSION.md`, `screenshots/`)
+
+If no recording is provided, generate a **synthetic happy path** using the data generation strategies in STAGE_TEST_REFERENCE.md (see "Synthetic Data Generation" section).
+
+## Step 1: Read Reference Materials
+
+Read these files to understand the testing infrastructure and patterns:
+
+1. `tests/e2e/docs/STAGE_TEST_REFERENCE.md` — what to test for each stage type, fixture availability, validation testing patterns
+2. `tests/e2e/fixtures/stage-fixture.ts` — available fixture methods and their signatures
+3. `tests/e2e/fixtures/interview-fixture.ts` — interview navigation fixture
+4. `tests/e2e/fixtures/protocol-fixture.ts` — protocol installation and network state inspection
+5. `tests/e2e/CLAUDE.md` — full e2e testing architecture guide
+6. `tests/e2e/specs/interview/silos-protocol.spec.ts` — reference test implementation to match style/structure
+7. `CLAUDE.md` — project coding conventions (path aliases, TypeScript, etc.)
+
+## Step 2: Extract and Analyze Protocol
+
+Extract the protocol JSON:
+
+```bash
+unzip -p "$1" protocol.json
+```
+
+From the extracted JSON, build a **stage map** — for each stage (by index), extract:
+
+- `type` — stage type (e.g., `NameGeneratorQuickAdd`, `EgoForm`, `Sociogram`)
+- `label` — display label
+- `subject` — `{ entity, type }` pointing to codebook entry (null for Information/Anonymisation)
+- `introductionPanel` — title and text (if present)
+- `form.fields[]` — array of `{ variable }` referencing codebook variables
+- `prompts[]` — array of prompt objects (with `createEdge`, `variable`, `highlight`, etc.)
+- `panels[]` — side panel configuration
+- `behaviours` — `maxNodes`, `minNodes`, `freeDraw`, etc.
+
+For each form field, resolve the variable UUID against the codebook:
+
+- Look up `codebook.[entity].[type].variables.[variableId]`
+- Extract: `name`, `type`, `component`, `validation`, `options`
+
+This gives you the field name (UUID), display name, input component type, validation rules, and available options for each form field.
+
+## Step 3: Analyze Recording (if provided)
+
+If `$2` is provided, read `$2/actions.jsonl` (one JSON object per line).
+
+Group actions by stage — track URL changes via the `step=N` query parameter. For each stage visited:
+
+- Extract the sequence of user actions (click, fill, press, select)
+- Note filled values and selected options
+- Note which nodes were created (names entered in quick-add or name generator forms)
+- Note edge-creating interactions (sociogram clicks, dyad census selections)
+
+The recording represents the **happy path** — the exact user journey to replay.
+
+### If no recording
+
+Generate a synthetic happy path from the protocol alone:
+
+1. Walk through all stages in order (index 0 to N)
+2. For each stage, use the **Synthetic Data Generation** section of STAGE_TEST_REFERENCE.md to determine what values to fill, how many nodes to create, etc.
+3. For conditional/skip logic, choose the path that visits the **most stages**
+4. Track synthetic state as you go — node names created on earlier stages are needed for bin/census/sociogram stages later
+
+## Step 4: Generate Test File
+
+Create `tests/e2e/specs/interview/<protocol-name>.spec.ts` where `<protocol-name>` is derived from the protocol name (kebab-case, lowercase).
+
+### File Structure
+
+Follow this exact pattern (from the reference implementation):
+
+```typescript
+/**
+ * <Protocol Name> Tests
+ *
+ * Tests interview stage navigation using a real .netcanvas protocol file.
+ */
+
+import path from 'node:path';
+import { expect, test } from '~/tests/e2e/fixtures/interview-test.js';
+import { expectURL } from '~/tests/e2e/helpers/expectations.js';
+
+const PROTOCOL_PATH = path.resolve(
+  import.meta.dirname,
+  '../../data/<filename>.netcanvas',
+);
+
+let sharedProtocolId: string;
+
+test.describe('<Protocol Name>', () => {
+  test.beforeAll(async ({ database, protocol }) => {
+    await database.restoreSnapshot();
+    const { protocolId } = await protocol.install(PROTOCOL_PATH);
+    sharedProtocolId = protocolId;
+  });
+
+  test.describe('Happy Path', () => {
+    test.describe.configure({ mode: 'serial' });
+
+    let interviewId: string;
+
+    test.beforeAll(async ({ protocol }) => {
+      interviewId = await protocol.createInterview(sharedProtocolId);
+    });
+
+    test.beforeEach(({ interview }) => {
+      interview.interviewId = interviewId;
+    });
+
+    test.afterEach(async ({ page, interview }) => {
+      const stepMatch = /step=(\d+)/.exec(page.url());
+      if (stepMatch?.[1]) {
+        const step = stepMatch[1];
+        // List stage indices with non-deterministic rendering
+        const highToleranceStages: string[] = [/* sociogram indices */];
+
+        await interview.capture(`stage-${step}-final`, {
+          maxDiffPixelRatio: highToleranceStages.includes(step)
+            ? 0.1
+            : undefined,
+        });
+      }
+    });
+
+    // One test() per stage...
+  });
+});
+```
+
+### Per-Stage Test Generation
+
+For each stage in the protocol, generate a `test()` block. Use the STAGE_TEST_REFERENCE.md to determine what to test.
+
+#### Mapping Recording Actions to Fixture Calls
+
+Translate recording actions to fixture method calls using these mappings:
+
+| Recording Pattern | Fixture Call |
+|---|---|
+| Navigate to URL with `step=N` | `interview.goto(N)` |
+| Click element matching next/forward button | `interview.nextButton.click()` |
+| Fill input within `[data-field-name="UUID"]` | `stage.form.fillText(UUID, value)` or `fillNumber`/`fillDate` based on codebook component |
+| Click radio within `[data-field-name="UUID"]` | `stage.form.selectRadio(UUID, optionLabel)` |
+| Click checkbox within `[data-field-name="UUID"]` | `stage.form.selectCheckbox(UUID, optionLabel)` |
+| Click toggle button within `[data-field-name="UUID"]` | `stage.form.selectToggleButton(UUID, optionLabel)` |
+| Fill quick-add input + press Enter | `stage.quickAdd.addNode(value)` |
+| Click "Add a person" button | `stage.nameGenerator.openAddForm()` |
+| Click "Finished" button in dialog | `stage.nameGenerator.submitForm()` |
+| Drag node from panel | `stage.nodePanel.dragNodeToMainList(label)` |
+| Click node on sociogram (connecting) | `stage.sociogram.connectNodes(from, to)` |
+| Drag node to ordinal bin | `stage.ordinalBin.dragNodeToBin(node, bin)` |
+| Drag node to categorical bin | `stage.categoricalBin.dragNodeToBin(node, bin)` |
+
+#### Determine Form Method from Codebook
+
+Use the codebook variable's `component` (or `type` if no component) to pick the right form fixture method:
+
+| Component | Method |
+|---|---|
+| `Text`, `TextArea` | `fillText` |
+| `Number` | `fillNumber` |
+| `DatePicker` | `fillDate` |
+| `RadioGroup` | `selectRadio` |
+| `LikertScale` | `selectLikert` |
+| `CheckboxGroup` | `selectCheckbox` |
+| `ToggleButtonGroup` | `selectToggleButton` |
+| `Boolean` | `selectRadio` (options are "Yes"/"No" or custom labels from codebook) |
+
+#### Comments
+
+Add a comment above each form field interaction with the field's display name and component type:
+
+```typescript
+// 1. Date of birth (DatePicker)
+await stage.form.fillDate('596c2ac2-...', '2000-06-15');
+
+// 2. Gender identity (RadioGroup)
+await stage.form.selectRadio('a06f06f5-...', 'Cisgender Male');
+```
+
+### Validation Tests
+
+For each form stage (EgoForm, AlterForm, AlterEdgeForm), examine the codebook variables for targeted validation rules. Generate validation test assertions **within the happy path test** for that stage:
+
+1. **Before filling fields**: Try to advance, verify validation blocks:
+   ```typescript
+   // Verify validation blocks advancement
+   await interview.nextButton.click();
+   await expectURL(page, /step=N/); // Still on same stage
+
+   // Verify required field errors
+   await expect(
+     stage.form.getFieldError('field-uuid'),
+   ).toBeVisible();
+   ```
+
+2. **Then fill fields normally** from the recording data.
+
+Only test these validations (skip others):
+- `required: true` — always test
+- `minValue` / `maxValue` — test if present
+- `minLength` / `maxLength` — test if present
+- `pattern` — test if present
+- `unique` — test if applicable (needs duplicate value scenario)
+- `sameAs` / `differentFrom` — test if present
+
+### Network State Verification
+
+The sync middleware uses a 3-second debounce with leading+trailing edges. Each `interview.goto()` destroys the current page, killing any pending trailing-edge syncs. Stages that set data used by downstream skip logic or filtering must explicitly wait for that data to persist.
+
+#### Form stages (EgoForm, AlterForm) must click Next to submit
+
+Form data lives in React Hook Form's local state until the form is submitted. **You must click `interview.nextButton` at the end of every form stage** to flush the data to Redux. Without this, the sync middleware never sees the data.
+
+For **EgoForm** stages, click Next as the last interaction (replaces the `toBeEnabled` assertion):
+
+```typescript
+// Submit form to flush data to Redux
+await interview.nextButton.click();
+```
+
+For **AlterForm** stages with slides, click Next after filling the **last slide** (the earlier slides already submit when you click Next to advance):
+
+```typescript
+// Submit last slide to flush form data to Redux
+await interview.nextButton.click();
+```
+
+Note: clicking Next navigates to the next stage, so the `afterEach` screenshot will capture the next stage's initial state rather than the current stage's final state.
+
+#### Persistence waits for skip logic
+
+After stages that set attributes consumed by downstream skip logic or filtering, add explicit waits using the protocol fixture. Available methods:
+
+- `protocol.waitForNodes(interviewId, expectedCount)` — after node creation stages
+- `protocol.waitForNode(interviewId, nodeName)` — when count alone is ambiguous
+- `protocol.waitForNodeAttribute(interviewId, nodeName, attributeId)` — after CategoricalBin, OrdinalBin, or AlterForm stages (checks for non-null value)
+- `protocol.waitForEgoAttribute(interviewId, attributeId, expectedValue)` — after EgoForm stages
+
+Example for a CategoricalBin stage with downstream skip logic:
+
+```typescript
+test('Stage N: CategoricalBin', async ({ interview, stage, protocol }) => {
+  await interview.goto(N);
+
+  await stage.categoricalBin.dragNodeToBin('Dan', 'Yes');
+  await stage.categoricalBin.dragNodeToBin('Alice', 'No');
+
+  await expect(interview.nextButton).toBeEnabled();
+
+  // Wait for the LAST categorized node's attribute to persist
+  await protocol.waitForNodeAttribute(
+    interview.interviewId,
+    'Alice',
+    'variable-uuid',
+  );
+});
+```
+
+**Always add `protocol` to the test's destructured fixtures** when using persistence waits.
+
+### Stages With Placeholder Fixtures
+
+Check the Fixture Availability Summary in STAGE_TEST_REFERENCE.md. If a stage type's fixture is marked **Placeholder**, generate a minimal test with a TODO referencing the placeholder:
+
+```typescript
+test('Stage N: Stage Label', async ({ page, interview }) => {
+  await interview.goto(N);
+
+  // TODO: stage.dyadCensus is a placeholder fixture — implement its
+  // interaction methods before writing full test assertions.
+  // See DyadCensusFixture JSDoc in stage-fixture.ts for the methods needed.
+  //
+  // Expected behavior from recording:
+  // - Dismiss intro panel
+  // - Select Yes/No for each node pair
+  // - Auto-advances after 350ms
+});
+```
+
+Always reference the `stage.<fixtureName>` property (e.g., `stage.dyadCensus`, `stage.narrative`) so the test structure is ready — it just needs the fixture methods implemented. Never use raw Playwright selectors as a fallback.
+
+### Skipped Stages
+
+If the recording skips certain stage indices (e.g., conditional stages), add a comment:
+
+```typescript
+// Stages N-M are skipped (conditional on <condition from codebook>)
+```
+
+### Browser-Specific Skips
+
+Add `test.skip()` for known browser limitations:
+
+```typescript
+// Skip geospatial on Firefox (no WebGL in Playwright's Firefox)
+test.skip(browserName === 'firefox', 'Firefox lacks WebGL support in Playwright');
+```
+
+## Step 5: Verify Protocol File Location
+
+Check if the `.netcanvas` file is already in `tests/e2e/data/`. If not, suggest copying it there and update the path constant accordingly.
+
+## Step 6: Output Summary
+
+After generating the test file, output:
+1. Path to the generated test file
+2. Number of stages covered
+3. Number of validation tests included
+4. List of stages with TODO placeholders (missing fixtures)
+5. Suggested next steps (copy protocol to test data, run tests, etc.)
+
+## Important Rules
+
+- **Always use path aliases** (`~/tests/e2e/...`) for imports, never relative paths
+- **Use `.js` extensions** in import paths (TypeScript with ESM)
+- **Field names are UUIDs** — always use the variable UUID from the codebook, not the display name
+- **Serial mode** — interview tests MUST use `test.describe.configure({ mode: 'serial' })`
+- **Soft assertions for screenshots** — the `afterEach` capture pattern handles this via `interview.capture()`
+- **No `console.log`** — project ESLint rule forbids it
+- **Follow existing patterns** — match the style, structure, and conventions of `silos-protocol.spec.ts` exactly

.dockerignore

@@ -4,6 +4,10 @@ node_modules
 npm-debug.log
 README.md
 .next
+.pnpm-store
+.e2e-assets
+.worktrees
+test-results
 .git
 .gitignore
 .env*

.env.development

@@ -0,0 +1,11 @@
+# Local development S3 storage — points at the MinIO container started by `pnpm dev`.
+# These values are intentional test credentials and are safe to commit.
+# To test the full setup wizard, comment out STORAGE_PROVIDER so the provider
+# is not pinned and the settings UI stays editable.
+STORAGE_PROVIDER=s3
+S3_ENDPOINT=http://localhost:9000
+S3_PUBLIC_URL=http://localhost:9000
+S3_BUCKET=fresco-dev
+S3_REGION=us-east-1
+S3_ACCESS_KEY_ID=minioadmin
+S3_SECRET_ACCESS_KEY=minioadmin