fix(usage): wire agent exclusions through usage filters

[prateek]

Fixes the pre-existing silent no-op where clicking an agent in Usage attribution updated local exclusion state, but the next usage requests did not send excludeAgent. The store now includes agent exclusions in the shared request params and counts them in hasActiveFilters, so the API sees the same filter state the UI shows.

The Usage toolbar also gets the Agent dropdown next to Project and Model. It uses the existing dropdown pattern, which makes agent exclusions visible and reversible without relying on attribution-row clicks. This stays scoped to Usage agent filtering and does not depend on the branch-dimension work.

Reviewers should start with frontend/src/lib/stores/usage.svelte.ts; UsagePage.svelte is the toolbar wiring, and the added component/store tests cover the request params plus the attribution-click path.

Fixes #976.

[roborev-ci]

roborev: Combined Review (fb5d625)

Medium issue found.

Medium

• frontend/src/lib/components/usage/UsagePage.svelte:177
  Agent exclusions are sent to the usage API, but /usage URL hydration still ignores exclude_agent and never clears usage.excludedAgents for deep links. Because excludedAgents is restored from localStorage, opening a shared URL with another filter, such as ?model=..., can silently keep excluding a previously hidden agent and show incorrect usage results.
  Fix: Add exclude_agent to the usage URL keys, hydrate usage.excludedAgents from params["exclude_agent"] ?? "", emit it from buildUsageUrlParams, and update the URL-param tests.

---

Reviewers: 2 done | Synthesis: codex, 6s | Total: 4m35s

[roborev-ci]

roborev: Combined Review (d24a85e)

No Medium, High, or Critical findings were reported.

Review 1 only reported a Low-severity URL persistence issue, which is omitted per instructions. Review 2 found no issues.

---

Reviewers: 2 done | Synthesis: codex, 7s | Total: 3m15s

[prateek]

Addressed the automated review finding: agent exclusions now round-trip through the URL (exclude_agent param), matching project exclusions, so a shared or reloaded /usage URL preserves them.

[roborev-ci]

roborev: Combined Review (55469df)

Summary verdict: One medium issue needs fixing before merge.

Medium

• frontend/src/lib/components/usage/UsagePage.svelte:52: The new Agent dropdown is backed by sessions.agents, but UsagePage never loads that metadata. On a direct /usage visit, the list can remain empty unless another component has already called sessions.loadAgents() or a sync invalidation happened. Load agent metadata when the usage page mounts, or make the dropdown trigger sessions.loadAgents() before opening.

---

Reviewers: 2 done | Synthesis: codex, 6s | Total: 3m26s

[prateek]

Addressed the round-2 finding: UsagePage now calls sessions.loadAgents() on mount, so the Agent dropdown is populated on a direct /usage visit instead of depending on the sidebar having loaded agent metadata first. Covered by a mount test.

[roborev-ci]

roborev: Combined Review (cd64f5f)

Medium issue found: active filter clearing does not clear agent exclusions.

Medium

• frontend/src/lib/components/usage/UsagePage.svelte:451
  Agent exclusions are now active usage filters, but SessionActiveFilters is still only wired for model/project usage filters. If a user excludes an agent and then uses the active-filters “Clear all” control, usage.excludedAgents remains set, so usage data stays filtered after the UI action implies all filters were cleared.

  Fix: Add agent filter props/handlers to SessionActiveFilters and pass usage.excludedAgents from UsagePage, including onRemoveAgent and onClearAgents={() => usage.selectAllAgents()} in the clear-all path.

---

Reviewers: 2 done | Synthesis: codex, 7s | Total: 5m23s

[prateek]

Addressed the round-3 finding: SessionActiveFilters gained an onClearAgents hook invoked by Clear all, and UsagePage wires it to usage.selectAllAgents(), so agent exclusions now reset with the rest. Kept to the clear-all hook to match how excluded projects are handled on this surface (they don't render chips here either); per-exclusion chips would be a separate change across both dimensions.

[roborev-ci]

roborev: Combined Review (cbe9a6a)

No issues found.

---

Reviewers: 2 done | Synthesis: codex | Total: 6m39s