Medical domains should be blocked.

Password manager domains should be blocked.

Test that invalid feed IDs don't affect cache key generation. This verifies the fix where we now use validated feed_ids consistently

Test that the get_river_cache_keys() helper method generates consistent cache keys regardless of feed list order. This is a

Large river feed lists that only differ after the first few ids must not share the same cache key.

Test that feed_stories() and truncate_river() use the same Redis pool. This verifies the fix where we changed truncate_river() to us

Categories should not count deleted archives.

Categories endpoint should require authentication.

Categories should return category counts.

A UC channel id with a cached feed title needs no channels.list call.

Without a stored feed title, fall back to the channels.list lookup.

last_1_min_time_taken should reflect the latest complete minute, not the oldest.

compute_story_score returns identical results whether the caller passes flat classifier lists or ScopedClassifiers indexes.

Positive (+1) still beats regular negative (-1).

Super downvote (-2) beats any positive classifier (+1).

compute_story_score should include author_regex score for Pro users

Two Stripe webhooks fire activate_premium in parallel for the same trial→paid conversion. Only one should send the new-premium email and

ES connection error should return empty list.

Empty content should be handled gracefully.

Content should survive compression/decompression roundtrip.

Unicode content should survive compression.

Conversations should be isolated per user.

Conversations list should require authentication.

Test creating an exact author classifier (is_regex=False)

Test creating a regex author classifier (is_regex=True)

Test creating an exact URL classifier (is_regex=False)

Test creating a regex URL classifier (is_regex=True)

Should be able to create a new conversation.

Custom section should use ES phrase search when available.

ES match already in top N should get its category reassigned to custom_X.

Custom section should fall back to keyword matching when ES not indexed.

Daily totals return subscription counts per day.

Stories sharing only a date should not cluster.

Verify we have a reasonable number of default blocked domains.

Verify we have blocked patterns for common sensitive pages.

Deleting a feed from 'Tech' should not affect 'Deep Tech'.

Story remains saved with empty user_tags list.

Delete should handle multiple archives.

Delete should handle nonexistent archive gracefully.

Delete should not delete other user's archives.

Delete endpoint should require authentication.

Delete should soft-delete a single archive.

Validation: can't delete empty tag name.

MStarredStoryCounts entry deleted.

Tag removed from user_tags list.

Ensure endpoint requires login.

Stories still exist as starred, just without the tag.

Detailed metrics include weighted score and raw counts.

Should not store when extension content is shorter.

Domain extraction should handle various URL formats.

Domains should include last_visit timestamp.

Domains endpoint should require authentication.

Domains should respect limit parameter.

Domains should return domain list with counts.

Empty feed_ids should return empty list without calling ES.