[#33] Add writer type filter and agent badge

[realproject7]

Summary

• Add writer filter (All / Human only / Agent only) to the discover page, wired to ?writer=all|human|agent query param
• Extract inline agent badge markup into a reusable AgentBadge component
• Filter Supabase queries by writer_type (0 = human, 1 = agent) on new/completed tabs; post-filter trending/rising results
• TabNav now preserves extra query params (writer filter persists across tab switches)

Fixes #33

Test plan

• Visit /discover — writer filter shows below tabs with "All" active by default
• Click "Human only" — URL updates to ?tab=new&writer=human, only writer_type=0 stories shown
• Click "Agent only" — URL updates to ?tab=new&writer=agent, only writer_type=1 stories shown
• Switch tabs while a writer filter is active — filter persists in URL
• Verify agent badge renders on StoryCard and story detail page for agent-written stories
• Verify no badge appears for human-written stories

🤖 Generated with Claude Code

[project7-interns]

Verdict: REQUEST CHANGES

Summary

The new/filter-completed paths are correct, and CI passed, but the trending/rising writer filter is implemented after ranking the global candidate set rather than in the underlying Supabase candidate query. That means writer=agent or writer=human can miss valid results and return an under-filled or incorrect top list.

Findings

• [medium] Trending/rising writer filtering happens too late, so filtered rankings are not actually correct
  • File: src/app/discover/page.tsx:113
  • Suggestion: Thread the writer filter into the ranking candidate fetch in lib/ranking.ts so the Supabase storyline query itself is constrained by writer_type before scoring/sorting. Filtering after getTrendingStorylines(..., 50) / getRisingStorylines(..., 50) only considers the global top 50 candidates, not the top writer-filtered set.

Decision

Request changes because issue #33 explicitly calls for filtering by writer_type, and the current trending/rising implementation can drop valid stories for the selected writer type.

[project7-interns]

T2b Review: REQUEST CHANGES

Agree with T2a — the post-filter on trending/rising is the blocking issue.

Blocking:
The filterByWriter(all, writer).slice(0, 20) pattern fetches 50 pre-ranked results then filters. If only 3 of 50 match the writer type, users see 3 results instead of 20, and valid stories ranked 51+ are invisible. Thread writerType into fetchCandidatesAndRatings in lib/ranking.ts so .eq("writer_type", ...) is applied before .limit(50).

Non-blocking:

• filterByWriter JSDoc says "client-side" but it runs in a Server Component — minor but misleading.

Everything else looks good:

• AgentBadge: clean, properly typed, reusable
• Query param handling: whitelist-validated, no injection risk
• WriterFilter + TabNav integration: correct URL construction, filter persists across tab switches
• Import paths all correct

[project7-interns]

Verdict: APPROVE

Summary

The follow-up fix resolves the correctness issue in the writer filter implementation by applying writer_type in the ranking candidate query before scoring and sorting. The discover-page filter, badge extraction, and query-param persistence now align with issue #33, and CI passed.

Findings

• None.

Decision

Approve because the implementation now satisfies the acceptance criteria end-to-end, stays focused, and the lint-and-typecheck check passed.

[project7-interns]

T2b Re-review: APPROVE

Fix is correct — writerType now threaded into fetchCandidatesAndRatings with .eq("writer_type", ...) applied before .limit(50). Post-filter removed entirely. Same pattern applied to new/completed tabs in queryTab. All 4 tabs now filter at the DB level. Ship it.