danvics/docling-studio
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 1

test(#256): Karate UI regression for Doc tab chunk mode

Browse source 
Setup via API (upload + analyse + wait COMPLETED), then drive the UI
to the chunks tab and assert the chunk list renders. Includes a direct
HTTP probe on /api/documents/{id}/chunks so an API-only regression
fails the test even if the UI somehow swallows it.

Tagged @critical so it runs in the smoke suite.
This commit is contained in:
Pier-Jean Malandrino 2026-05-07 11:09:19 +02:00
parent 4a422a8464
commit 8ac81b9461
1 changed files with 53 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

53
e2e/ui/src/test/resources/documents/doc-tab-chunk-mode.feature Normal file
View file

@ -0,0 +1,53 @@
@ui @critical
Feature: UI — Doc tab chunk mode renders canonical chunks (#256)
# Regression coverage for the bug where opening the chunks tab on a
# document returned 404 because /api/documents/{id}/chunks was not
# implemented backend-side. Setup is API-driven (upload + analyse) so
# the test is fast and deterministic; the UI portion only exercises
# navigation + chunk rendering, not the parse pipeline.
Background:
* url baseUrl
Scenario: Open the chunks tab and see the canonical chunkset (no 404)
# 1. Setup via API: upload + run an analysis (with chunking) and wait
# until the doc-centric chunkset gets promoted by the analysis hook.
* def upload = call read('classpath:common/helpers/upload.feature') { file: 'small.pdf' }
* def docId = upload.docId
Given url baseUrl
And path '/api/analyses'
And request { documentId: '#(docId)', chunkingOptions: { chunkerType: 'hybrid', maxTokens: 256, mergePeers: true, repeatTableHeader: true } }
When method POST
Then status 200
* def analysisId = response.id
# Poll until the analysis is COMPLETED (chunks get promoted in the
# same step via AnalysisService → ChunkService.promote_from_analysis).
Given url baseUrl
And path '/api/analyses', analysisId
And retry until response.status == 'COMPLETED' || response.status == 'FAILED'
When method GET
Then status 200
And match response.status == 'COMPLETED'
# 2. Sanity-check the new endpoint over HTTP — proves the 404 is gone.
Given url baseUrl
And path '/api/documents', docId, 'chunks'
When method GET
Then status 200
And assert karate.sizeOf(response) > 0
# 3. Drive the UI: open the workspace page and click the chunks tab.
* driver uiBaseUrl + '/docs/' + docId
* waitFor('[data-e2e=tab-strip]')
* click('[data-e2e=tab-chunks]')
# 4. Verify the chunks list rendered (no 404 banner, chunk-list visible).
* waitFor('[data-e2e=chunks-editor]')
* waitFor('[data-e2e=chunk-list]')
* assert karate.sizeOf(locateAll('[data-e2e=chunk-list] .chunk-card')) > 0
# 5. Cleanup via API.
* call read('classpath:common/helpers/cleanup-by-name.feature') { filename: 'small.pdf' }