Repository navigation
Description
Problem or use case
Orca’s AI Provider Accounts / status-bar rate-limit strip already tracks Claude, Codex, Gemini, Antigravity, OpenCode Go, Kimi, MiniMax, and Grok. Several API gateways that Orca users (and OMP panes inside Orca) already call are missing:
| Provider | Role for Orca users | Gap today |
|---|---|---|
| OpenRouter | Multi-model API gateway used from OMP and other agents | No Accounts row, no status-bar chip |
| TokenRouter | OpenAI-compatible router with per-key quotas + wallet | Same |
| Ollama Cloud | Hosted cloud models (ollama-cloud/* in OMP) |
Same; users leave Orca to check ollama.com/settings/usage |
This is the same class of request as Fireworks (#20812), DeepSeek (#12869), Cursor (#15516), Copilot (#6892).
Out of scope for this issue: making these first-class agent launch targets. This request is only live usage / quota in Accounts + the status bar.
Proposed solution
Add three optional usage providers, following MiniMax / Grok / Fireworks patterns:
- Credentials in Settings → Accounts (desktop-safe storage; mirror MiniMax API-key paste)
- Fetchers under the existing rate-limit pipeline that map vendor payloads into
ProviderRateLimits(session/weekly/monthlyand/or a money/credit balance readout where % windows do not apply — same display path Fireworks proposes) - Status-bar chips + visibility toggles in Appearance
unavailable= not configured; exhausted-but-configured balances stay visible- Focused unit tests per fetcher
Confirmed vendor endpoints (probed 2026-09-18)
OpenRouter — API key (OPENROUTER_API_KEY / paste sk-or-…):
GET https://openrouter.ai/api/v1/auth/key→ per-keyusage_*, optionallimit/limit_remaining,free_model_daily_requests.{used,limit,remaining}GET https://openrouter.ai/api/v1/credits→{ total_credits, total_usage }(remaining ≈ credits − usage)
Suggested UI:
- Primary: remaining credit balance (money), analogous to Fireworks
- Secondary when present: free-model daily request window as a % bar; optional key spend limit when
limitis set
TokenRouter — management key preferred; API key as fallback:
- Management:
GET https://api.tokenrouter.com/api/management/self/wallet→topUpBalance,toppedUpSpent, vouchers (Management API docs) - Management:
GET https://api.tokenrouter.com/api/management/api-keys→ per-keyremain_quota/used_quota/unlimited_quota - API key fallback:
GET https://api.tokenrouter.com/v1/dashboard/billing/usage+…/subscription(OpenAI-shaped; soft/hard limits are not always meaningful)
Suggested UI:
- Primary: wallet
topUpBalance(money remaining) - Optional secondary: selected API key remain/used quota when not
unlimited_quota
Ollama Cloud — API key (OLLAMA_CLOUD_API_KEY / paste from ollama.com/settings/keys):
GET https://ollama.com/api/usage→
limits.session.usageandlimits.weekly.usage(0–1 fractions in practice) plus per-model request counts and period activity cost
Suggested UI:
- Best fit for existing session + weekly % bars (same shape as Claude/Codex chips)
- Optional activity cost under the row
Implementation slices (reviewable PRs)
- Ollama Cloud first — maps cleanly onto existing
session/weeklywindows; smallest UI delta - OpenRouter — credit balance + optional free-tier daily window (Fireworks-like money readout)
- TokenRouter — management-key wallet + optional per-key quota
Shared plumbing per provider (same checklist as #12869):
- Extend rate-limit state /
ProviderRateLimits['provider'] - Credential storage + Accounts pane row
- Fetcher + unit tests
- Status-bar roster / visibility wiring
- i18n strings
Alternatives or additional context
- Workaround today: check each vendor dashboard; OMP can use the models, but Orca’s quota strip ignores them.
- Prefer documented HTTP APIs + user-pasted keys over scraping dashboards or reading another app’s keychain (MiniMax/Grok-style paste is fine).
- Do not invent % windows when the vendor only exposes a dollar/credit balance — show money remaining (Fireworks [Feature]: Fireworks.ai usage in the status bar (rate-limit / usage provider) #20812 precedent).
- Cross-platform: macOS / Linux / Windows; remote runtimes should degrade to “not available” when no local credential exists (same rule as Cursor [Feature]: Add Cursor usage tracking (status-bar rate limits + Accounts pane) #15516).
Related
- [Feature]: Fireworks.ai usage in the status bar (rate-limit / usage provider) #20812 Fireworks usage
- [Feature]: DeepSeek usage in the status bar (rate-limit / usage provider) #12869 DeepSeek usage
- [Feature]: Add Cursor usage tracking (status-bar rate limits + Accounts pane) #15516 Cursor usage
- [Feature]: Add GitHub Copilot usage tracking (status-bar rate limits + Accounts pane) #6892 Copilot usage
- MiniMax / Grok Accounts patterns already in-tree
I can implement these as focused PRs (Ollama Cloud → OpenRouter → TokenRouter) after maintainers confirm the credential + display shapes above.
Activity
jamesjfoong commentedon Nov 3, 2026
Ollama Cloud implemented in PR #37829: Ollama joins the rate-limit providers (documented
balance+30d usageendpoints, sealed OLLAMA_API_KEY entry in Settings ▸ Accounts, status-bar bar + letter). OpenRouter/TokenRouter remain open in this issue — happy to follow the same pattern for them if the approach lands well.