feat(tui): hide auto-design-router noise from queue by default

[cpcloud]

Summary

• Adds show_classify_jobs config (off by default, global + per-repo *bool override) that hides auto-design-router byproducts from the TUI queue: rows with job_type='classify' and rows with status='skipped'. Both predicates are scoped to source='auto_design' so a future pipeline that adopts the same job_type or status for unrelated reasons isn't silently swallowed.
• Server-side filtering via new /api/jobs?hide_classify_jobs=true query param backed by storage.WithHideClassifyJobs. SQL uses COALESCE on j.source so rows with source IS NULL aren't dropped by NULL three-valued logic.
• Per-repo override applies only when the queue is filtered to a single repo, otherwise the global value applies, avoiding one repo's preference leaking into a multi-repo view.
• Session-only s hotkey toggles visibility; the queue footer label and the ? help view both reflect the current state.
• Hitting l on an auto-design classify or terminal skipped row renders the classifier verdict, skip_reason, and any internal error in a header band above the (typically empty) streamed log. Reasoning lines are width-truncated and have embedded newlines/tabs folded to spaces so each occupies exactly one terminal row, keeping logVisibleLines accurate. Verdict text branches on job.Status so failed/canceled classifiers don't claim "in progress".
• Effective show-classify value is cached on the model and recomputed only on filter or session-toggle changes, keeping render and fetch paths free of LoadRepoConfig disk I/O.

[roborev-ci]

roborev: Combined Review (2a5fce6)

Medium issue found: the TUI log header will not show auto-design classifier verdict/reason data because list job records do not populate the required fields.

Medium

• cmd/roborev/tui/render_log.go:137
  The new log header path depends on job.Source == "auto_design", but log jobs come from m.jobs, populated via ListJobs. That query does not select/scan j.source or j.skip_reason, so real classify/skipped rows have empty Source, and the new verdict/reason lines never render.
  Fix: Include j.source and j.skip_reason in ListJobs’ SELECT/scan path, or load full job details before rendering classify log headers.

---

Synthesized from 3 reviews (agents: codex, gemini | types: default, security)

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 66.25000% with 54 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
internal/daemon_client/client.gen.go	0.00%	21 Missing ⚠️
cmd/roborev/tui/handlers_queue.go	0.00%	10 Missing ⚠️
cmd/roborev/tui/render_log.go	83.60%	8 Missing and 2 partials ⚠️
cmd/roborev/tui/handlers_msg.go	33.33%	3 Missing and 1 partial ⚠️
cmd/roborev/tui/render_queue.go	50.00%	1 Missing and 2 partials ⚠️
cmd/roborev/tui/fetch.go	71.42%	2 Missing ⚠️
cmd/roborev/tui/handlers.go	0.00%	2 Missing ⚠️
cmd/roborev/tui/tui.go	90.00%	2 Missing ⚠️

📢 Thoughts on this report? Let us know!

[roborev-ci]

roborev: Combined Review (feda678)

Review verdict: one medium issue should be addressed before merge.

Medium

• cmd/roborev/tui/render_log.go:148 - Skipped auto-design rows always render as Auto-design verdict: no design review needed, but classifier execution failures are also converted into skipped rows with job.Error populated. Those rows did not produce a clean “no design review needed” verdict, so the log view can mislead operators about a classifier failure. Distinguish job.Error != "" in the skipped auto-design branch and render a classifier failure or degraded-skip message instead.

---

Synthesized from 3 reviews (agents: codex, gemini | types: default, security)

[roborev-ci]

roborev: Combined Review (3a86be2)

High: TUI queue can break when listing standard review jobs due to NULL string scans.

High

• internal/storage/jobs.go around line 846: ListJobs selects nullable j.skip_reason and j.source directly, then scans them into string fields. For standard review jobs where these columns can be NULL, this can fail with sql: Scan error... converting NULL to string is unsupported, breaking the TUI queue. Use COALESCE(j.skip_reason, '') and COALESCE(j.source, '') in the query.

---

Synthesized from 3 reviews (agents: codex, gemini | types: default, security)

[cpcloud]

False positive on the latest roborev-ci finding. The review claims ListJobs scans nullable j.skip_reason and j.source directly into string fields, which would fail on NULL.

In fact the scan targets are sql.NullString:

// internal/storage/hydration.go
type reviewJobScanFields struct {
    ...
    SkipReason        sql.NullString
    Source            sql.NullString
}

The call site at internal/storage/jobs.go:880-887 scans into &fields.SkipReason, &fields.Source — sql.NullString handles NULL natively. The hydration step then unwraps via .Valid:

if fields.SkipReason.Valid {
    job.SkipReason = fields.SkipReason.String
}
if fields.Source.Valid {
    job.Source = fields.Source.String
}

TestListJobsWithHideClassifyJobs already exercises this with a row whose source IS NULL (the test asserts that row remains visible after the hide filter); it passes. No code change.