Should generate variants with both http and https.

Should generate variants with and without trailing slash.

Should generate variant without www when original has it.

Should generate variants with and without www.

get_archive_content should handle archive not found.

get_archive_content should return archive content via execute_tool.

get_archive_content tool should be defined.

get_archive_summary tool should be defined.

Get blocklist should return custom user settings.

Get blocklist should return default lists.

When folder_feed_ids is passed to get_classifiers_for_user, folder-scoped classifiers that don't apply to the current feed should be

Test that get_classifiers_for_user splits authors and author_regex

Test that get_classifiers_for_user includes URL classifiers

InvalidSchema (WhatsApp link etc.) should not crash.

MissingSchema (user passes search query instead of URL) should not crash.

When response has no Content-Type header, should not raise TypeError.

ReadTimeout on JSON feed check should not crash.

A proxied openrss URL must resolve to the proxy feed, never a rewritten gdata feed.

Adding an openrss preview URL must resolve to the /feed/ feed, not the preview.

Subscribing via a preview URL reuses an existing legacy feed, not a duplicate.

get_or_create should return existing settings.

get_or_create should create settings for new user.

get_suggested_questions should return a list.

Globally scoped classifiers match regardless of feed.

Global query should also support phrase search.

Empty URL should return empty string.

Invalid URLs should be handled gracefully.

Current yt3.googleusercontent.com avatars should not be treated as generic.

Image classifier notifications match the same way as text prompts.

Muted feeds don't count against the Premium limit, so they don't make a user mega.

Verify the query includes a date range filter.

Blocked URL should return blocked response.

Valid ingest should create archive.

Deleted archive should not appear in default list.

Ingesting same URL twice should update, not create.

Missing title should return error.

Missing URL should return error.

Ingest endpoint should require authentication.

Ingest response should include archive_id.

Archives should be exportable as CSV.

Archives should be exportable as JSON.

Archive created via ingest should appear in list.

Content ingested should be retrievable.

Revisiting URL should update archive fields.

Ingest should increment total_archived.

All optional fields should be stored.

Content should be stored when provided.

Archive should be created without content.

Invalid feed IDs don't crash or pollute data.

Invalid story hashes and short reads don't crash or pollute data.

Test that river loading does NOT use ZUNIONSTORE for aggregation. The lazy merge path should avoid the blocking ZUNIONSTORE entirely.

Test that lazy merge pagination works: page 2 extends the cache from page 1 rather than replacing it, so subsequent pages load correc

Test that marking stories as read between page loads doesn't cause stories to be skipped on subsequent pages (read_filter='unread').

A folder river's cache key must stay anchored to the validated requested feed set even when one feed reaches zero unread stories betw

Test that lazy merge (k-way heap) returns stories in correct newest-first order. All river loads now use lazy merge instead of ZUNION

Test that lazy merge works with oldest-first ordering.