feat: add Codex quota usage indicator

[safzanpirani]

Summary

• add a shared Codex quota usage fetcher for ChatGPT's wham usage endpoint
• add /codex-quota with /quota alias for primary and secondary quota windows
• show Codex quota pace bars in the CLI status bar for openai-codex sessions
• add configurable status-bar detail levels via display.codex_quota

Configuration

The status-bar indicator defaults to the compact/minimal view:

display:
  codex_quota: minimal

Available values:

• minimal — default. Shows compact quota pace bars plus used percentages.
• detailed — shows quota pace bars plus reset countdown/clock metadata, e.g. 2%·4h39m→1:49pm.

Switch with:

hermes config set display.codex_quota detailed
hermes config set display.codex_quota minimal

The slash command output remains detailed regardless of this setting:

/codex-quota
/quota

Minimal preview

Detailed preview

Test Plan

• venv/bin/python -m pytest tests/test_codex_quota.py tests/hermes_cli/test_commands.py tests/cli/test_cli_status_bar.py tests/hermes_cli/test_config.py -q -o 'addopts='

[teknium1]

Thanks for the Codex quota status-bar work. Current main now has a shared Codex account-usage client, so this needs a focused salvage rather than a parallel fetcher.

Problems

• agent/codex_quota.py:261 broadly catches resolver failures and then selects a pooled credential. A transient refresh failure can therefore show limits for another account. Main explicitly prevents that in agent/account_usage.py:459-464; the guard test is tests/agent/test_account_usage.py:121-152 (130e2337c24810bc0afa793495795b832cf593be).
• Current CLI /usage already passes live provider/base-URL/API-key context into the shared fetcher at cli.py:9681-9698. The proposed separate client should not bypass that established resolution path.

Suggested changes

• Build the status-bar view from agent.account_usage.fetch_account_usage and retain its fail-open credential handling.
• Add status-bar coverage in tests/cli/test_cli_status_bar.py, including non-Codex and transient-credential-failure cases.

Automated hermes-sweeper review.

[teknium1]

Do not catch every resolver failure and fall back to pool.select(): a transient refresh/network failure can make this display quota for a different ChatGPT account. Current main fixed this exact class by limiting pool fallback to AuthError in agent/account_usage.py:459-464 (guarded by tests/agent/test_account_usage.py:121-152).

[GottZ]

This was generated by AI during triage.

Summary

Eight PRs address or reference Codex quota visibility, but no Verify verdicts are recorded; the supplied diffs range from direct status-bar implementations to broader or superseded account-usage work. The strongest current fit is #44506, which reuses the shared account-usage fetcher and adds a scoped, nonblocking Codex session-limit segment, whereas several alternatives introduce parallel fetch paths or are based on obsolete status-bar code.

Related pull requests

• #6250 [closed] related — (+6189/-333) — do not consolidate: although titled as Codex rate-limit status-bar work, the supplied broad diff mixes credential-pool, vision-provider, worktree-cleanup, plugin, and other unrelated changes, so it is not an isolated fix for the quota-visibility cause; its closed state remains relevant because this high-risk patch is not a focused implementation.
• #11177 [closed] related — (+539/-3) — superseded reference: this closed PR added a Codex-only wham/usage client and exposed quota windows through CLI and Gateway /usage, not a focused current status-bar implementation. Contributor review records it as superseded by merged #13428, whose documented relevance is the broader three-provider implementation; the supplied evidence does not establish lineage from #13428 to the current agent.account_usage code.
• #19063 related — (+625/-1) — salvage only, not merge as-is: it adds quota commands, configurable pace bars, and a new agent/codex_quota.py, but its broad resolver exception can fall through to a different pooled account and it duplicates the established account-usage fetch path. Despite the keep_open review on #19063, the diff shows the specifically reviewed cross-account fallback hazard, so only its UI/formatting concepts should be retained unless it is rebuilt on agent.account_usage.fetch_account_usage with transient-failure coverage.
• #20493 [closed] related — (+299/-20) — duplicate implementation: this closed PR adds cached 5h/weekly status-bar bars by spawning codex app-server, creating a second quota acquisition path instead of reusing shared account usage. It remains relevant as the original version of the same-author re-submission #20953.
• #20786 [closed] duplicate — (+125/-0) — useful shared-path reference: this closed PR correctly builds a compact asynchronous label on agent.account_usage.fetch_account_usage, directly addressing status-bar visibility without a parallel client. It remains relevant because its implementation is the closest architectural precursor, but contributor discussion closed it as a duplicate of #19063 and its display is limited to one terse primary/session percentage.
• #20953 [closed] related — (+300/-20) — close/keep closed as duplicate: it is the rebased re-submission of #20493, with the same codex app-server JSON-RPC probe and status-bar formatting. Contributor discussion explicitly identifies it as a duplicate of #20493 and overlapping #19063 and #20786.
• #33836 related — (+207/-0) — broader salvage, not the focused merge candidate: it generalizes an opt-in badge across Codex, Anthropic, and OpenRouter using agent.account_usage, but the diff targets an old status-bar/config layout and lacks tests for single-flight refresh, retry, and fetch exceptions. Despite the keep_open review on #33836, those diff-level integration and cache-coverage gaps make it less merge-ready than #44506; its multi-provider scope can be pursued separately after rebasing.
• #44506 related — (+581/-0) — preferred merge candidate after required approval: the current diff adds a wide-tier Codex session percentage/reset segment, performs no render-thread network I/O, uses agent.account_usage.fetch_account_usage, scopes cached results to an immutable runtime-client credential fingerprint, and tests provider switches, credential failover, single-flight refresh, width fallback, and formatting. This aligns with the keep_open review on #44506, and the subsequent update directly addresses that review's unsafe-fallback objection without introducing a separate quota resolver.

Duplicates

#20493 and #20953 are effectively the same codex app-server status-bar implementation; #19063, #20786, #33836, and #44506 overlap on asynchronously surfacing account quota in the status bar, but differ materially in fetch-path safety, provider scope, and UI detail.

Suggested consolidation

Merge #44506 after its pending CI/maintainer approval because its supplied current diff is the most focused implementation and explicitly protects provider and credential scope while reusing shared account usage. Keep #11177 and #20786 closed as superseded/reference work, keep #20493 and #20953 closed as duplicates, and do not revive #6250. Close #19063 as superseded by #44506 despite its keep_open review because its present diff retains the reviewed parallel-fetch and cross-account fallback defect; close or retarget #33836 despite its keep_open review because its broader multi-provider idea is separable, while its supplied diff is stale and does not test the cache/failure paths required by that review.

Complex graph

flowchart LR
    classDef open fill:#dbeafe,stroke:#1d4ed8,color:#1e3a8a
    classDef merged fill:#dcfce7,stroke:#15803d,color:#14532d
    classDef closed fill:#e5e7eb,stroke:#6b7280,color:#1f2937
    classDef unverified fill:#f3f4f6,stroke:#9ca3af,color:#374151
    classDef best stroke-width:3px,stroke:#b45309
    classDef target stroke-width:3px,stroke:#4338ca
    subgraph Dup19063 ["PRs duplicating each other"]
        P19063["PR #19063 (open)"]
        P20786["PR #20786 (closed)"]
    end
    class P19063 open
    class P20786 closed
    class P19063 target
    click P19063 "https://github.com/NousResearch/hermes-agent/pull/19063"
    click P20786 "https://github.com/NousResearch/hermes-agent/pull/20786"

Loading

Graph: solid arrow = fixes / best fix, dashed arrow = partial or unverified (see edge label); boxed group = PRs duplicating each other; amber border = best fix; indigo border = target; gray node = closed (state tag in the node label).

Cross-PR triage: Reviewed 8 pull requests and 0 issues in this complex. Each diff was read against this issue; Assessment working set: 140 kB of PR diffs, 13 kB of issue/PR text, 7 kB of discussion (9 comments), 0 verify verdicts. verdicts reflect diff content, not PR titles. Part of an automated triage batch.