# Changelog

## 0.1.39 (2026-09-12)

### Fixed

- **Guarded footer render against an uninitialized theme proxy** (pi-budget
  parity): `renderSubscriptionLine` dereferenced `ctx.ui.theme.fg` unguarded —
  if the theme isn't ready yet the throw escapes as a rejected promise and can
  exit pi (the same unhandledRejection class 0.1.37/0.1.38 fixed elsewhere).
  The footer is now best-effort: skipped when the theme isn't available.
- **Finite-cost guard on `message_end` accumulation** (pi-budget parity): a
  string or NaN `cost.total` previously hit `+=` directly — a string cost
  concatenated onto the accumulator and garbled every subsequent footer.
  Costs are now coerced with `Number()` and only finite positive values
  accumulate.
- README intro: Router (pi-router) listed among supported providers.

## 0.1.38 (2026-09-07)

### Fixed

- **Crash (pi exits) — third arm of the stale-ctx class**: pi 0.85.1 can
  invalidate the extension ctx without ever delivering a matching
  `session_shutdown` (orphaned/replaced runtime teardown; extension instances
  are shared across sessions), so the 60s usage-refresh interval can fire with
  `state.ctx` still installed but stale — the 0.1.18/0.1.37 guards (fire-time
  ctx resolution, identity-guarded shutdown) never see it. `refreshUsage` is
  async, so its throw became a rejected promise discarded by `void` →
  `unhandledRejection` → pi's `uncaughtException` handler → exit. All deferred
  refresh call sites now go through `deferRefresh`, which catches the
  rejection and self-disarms (stops timers, drops in-flight state and the
  stale ctx); `session_start` re-arms with the fresh ctx. Same crash class as
  pi-messenger#25. The `/sub` command path disarms the same way when run
  against an orphaned stale ctx, and regression tests now cover the 60s
  interval arm, the command arm, and matcher independence from pi's exact
  error wording.

## 0.1.37 (2026-09-07)

### Fixed

- **Legible provider API errors**: structured API failures (e.g. Z.ai's
  `500 Internal service error` returned by `api.z.ai/api/monitor/usage/quota/limit`
  during its 2026-09-07 outage) no longer masquerade as a generic
  "usage unavailable" — the footer shows the server's own message
  (`Sub Z.ai (Anthropic) API error: Internal service error`), with
  credential-shaped material (`sk-…`, `Bearer …`, JWTs) scrubbed and length
  capped; auth-looking messages stay redacted.
- **Crash (pi exits) on session replacement (`/new`, fork, switch, `/reload`)**:
  a usage-refresh debounce timer armed by a late `after_provider_response` event
  — delivered after `session_shutdown` had already cleared the previous timers —
  captured the old extension ctx; ~2s later the timer fired, touched the now
  stale `ctx.ui`, and the uncaught error killed pi. Deferred helpers
  (`renderSubscriptionLine`, `refreshUsage`, `startTimer`, `scheduleRefresh`,
  `updateActiveAdapter`) no longer take a captured ctx: they resolve `state.ctx`
  at execution time. `state.ctx` is installed only by `session_start` (fires
  before a session's other events, always fresh) and cleared by
  `session_shutdown`, so mid-session handlers can never reinstall a stale ctx
  and timers firing in the teardown window safely no-op. `session_shutdown`
  now no-ops entirely unless it belongs to the installed session
  (`state.ctx === ctx`), so a late old-session shutdown delivered after the
  next `session_start` can neither touch an invalidated ctx nor stop the live
  session's refresh timer / drop its in-flight fetch. Regression tests in
  `extensions/test/stale-ctx-regression.test.ts`.

### Changed

- **Tests are now gated**: `npm test` runs both test files via
  `node --import tsx --test` (new `tsx` devDependency), and CI runs
  `npm ci && npm test` for pi-sub instead of only `npm pack --dry-run` — the
  stale-ctx regression tests can no longer be bypassed by a refactor.

## 0.1.36 (2026-09-07)

### Fixed

- **Removed the incorrect dead `tok-per-sec` module** (added in 0.1.35, never
  wired in): its `withThinking` formula summed `reasoning` on top of `output`,
  but Pi's `usage.output` already includes reasoning tokens (`reasoning` is a
  subset), which would have inflated thinking-mode tok/s up to ~2×. The live
  footer/footer tok/s math (`output / elapsed`) was and remains correct in both
  thinking and normal mode.

### Added

- **`/sub` thinking/answer split** — when the model reasoned, the details line
  now reads e.g. `Last response: 46 tok/s (36 think + 10 answer)` instead of a
  bare total, using the correct subset math (`answer = output − reasoning`).
  Footer keeps the single total number.

## 0.1.35 (2026-09-05)

### Added

- **`zai-anthropic` provider usage tracking** — GLM through the Anthropic-compatible endpoint (`api.z.ai/api/anthropic`, registered by pi-model-tools) now shows the same quota footer as `zai`/`zai-coding-cn`: 5-hour/weekly windows, MCP monthly allowance, and model/tool breakdowns, keyed by the auth.json `zai-anthropic` credential, labeled `Z.ai (Anthropic)`.

## 0.1.34 (2026-09-05)

- Widen Pi SDK peer range to `>=0.80.8 <0.86.0` for Pi 0.85.0 compatibility (no breaking changes; peer cap widening only).

## 0.1.33 (2026-08-29)

### Added

- `/sub` argument completion offers `refresh`.

## 0.1.32 (2026-08-22)

- **DeepSeek via OmniRoute now shows the real USD balance** (e.g. `M:$18.25`)
  instead of the misleading aggregate windows (R:100%/5H W:0%/2D). Credit-based
  upstreams report "Unavailable" session/weekly in the usage text; pi-sub now
  recognizes that shape and fetches `credits_usd` from the management usage
  API — grant the router key the `manage` scope in the OmniRoute dashboard
  (done), or set `ROUTER_MGMT_TOKEN`. Without it, the footer degrades to the
  clean endpoint display (no aggregate leak).
- The plain-retry fallback (drop `?provider=`) now fires only on "No cached
  usage data" (wrong/unknown slug), not on "Unavailable" windows.
- pi-sub now reads `.env.local`/`.env` (cwd then `~/.pi/agent`) for its own
  env vars — same discovery chain as pi-munin, stdlib parser, no deps.

## 0.1.31 (2026-08-22)

- **Fixed glm-cn quota via OmniRoute**: OmniRoute's connection slug for the
  Z.ai (CN) upstream is `glm-cn`, not `zai-coding-cn` (a Pi provider id).
  Wrong slug returned "No cached usage data", so the plain-retry fallback
  showed the aggregate snapshot (e.g. R:100% W:0%) instead of the Z.ai windows.
  Alias `glmcn` now maps to the correct slug; footer shows `R:99%/3H`-style
  windows matching the direct Z.ai display.
- OmniRoute "reset in 2h 55m" countdowns are now converted into compact
  footer labels (`/3H`), and "Unavailable" windows are skipped (previously
  `0%`-style aggregates leaked in via the fallback).

## 0.1.30 (2026-08-22)

- Router usage is now **provider-scoped**: the active router model's upstream
  prefix (e.g. `command-code` from `router/command-code/deepseek/deepseek-v4-flash`)
  is passed as `?provider=` to the OmniRoute usage API, so the footer shows that
  upstream's quota (e.g. opencode-go session/weekly windows).
- Aliases normalized before the API call: `cmd`→`command-code`, `oc`→`opencode-go`,
  `ds`→`deepseek`, `glmcn`/`glm-cn`→`zai-coding-cn`; generic router aliases
  (auto/aug/no-think/tllm/combo/openrouter/nvidia/…) are filtered out.
- When the selected upstream has no cached quota ("No cached usage data"),
  the adapter retries without `?provider=` so the footer still shows the best
  snapshot instead of empty windows.
- Switching between router models with different upstreams now refreshes
  immediately (adapter id includes the prefix, so stale in-flight results are
  discarded via the generation guard).
- `/sub` breakdown shows only the Provider quota section — personal USD budget
  lines are no longer surfaced.
- Self-check extended (17 assertions) and now exercises the real
  `routerUpstreamPrefix` (aliases + generic filter).

## 0.1.27 (2026-08-22)

- Router adapter now fetches real usage from OmniRoute instances: GET
  `<origin>/api/usage/om-usage` (Bearer = router API key from auth.json or
  ROUTER_API_KEY env). The plain-text report (Personal quota Daily/Weekly +
  Provider quota Session/Weekly) is parsed into the footer R:/W: windows.
  Non-OmniRoute routers 404 → fall back to the endpoint-only display.
  When the per-key usage command is disabled, the footer shows a hint to
  enable it in the OmniRoute dashboard (API Keys → the key).
- New pure parser `parseOmniUsageText` with an env-gated self-check
  (PI_SUB_SELF_CHECK=1) — pi-sub stays pack-only for CI.

## 0.1.26 (2026-08-21)

- Router adapter: reads `router.baseUrl` from `~/.pi/agent/settings.json`
  (env `ROUTER_BASE_URL`/`NINE_ROUTER_BASE_URL`) now that pi-router moved the
  endpoint URL out of the old `9router-config.json` (which is migrated away on
  pi-router load). Footer shows for both new `router/` and legacy `9router/`
  provider models.

# Changelog

## 0.1.33 (2026-08-29)

### Added

- `/sub` argument completion offers `refresh`.

## 0.1.29 (2026-08-15)

### Features

- **tok/s speed now shown for all providers.** Previously the footer was cleared
  for providers without a subscription adapter (e.g. `ollama`); now the last
  response speed renders as a standalone `145 tok/s` line, and `/sub` reports
  provider/model plus last and session-average speed instead of only
  "tracking inactive".

## 0.1.28 (2026-08-09)

### Improvements

- **Command Code monthly balance now shown in the footer.** The footer adds a
  compact `M:$X.XX` segment (monthly credit balance from
  `/alpha/billing/credits`) next to the 5-hour/weekly windows, e.g.
  `(Command Code key#…) R:100%/5H W:100%/7D M:$69.99`. The `/sub` detail view
  keeps the `Monthly: $X remaining` breakdown line.

## 0.1.27 (2026-08-09)

### Features

- **Command Code now shows live usage windows.** The adapter fetches
  `https://api.commandcode.ai/alpha/billing/credits` with the same Provider
  API key used for `/provider/v1` models and renders the 5-hour and weekly
  rolling windows (USD used/cap, remaining% + reset countdown) plus a
  `Monthly: $X remaining` balance line — matching what the `cmd /usage` CLI
  shows. Previously the footer showed session cost only.

## 0.1.26 (2026-08-09)

### Features

- Added Command Code (`commandcode`) provider support. When a `commandcode`
  model is active, the footer shows the account label plus session cost and
  tok/s, and `/sub` reports the provider. Command Code's Provider API does not
  expose usage windows via its API key (the rolling-window data is
  web-session-cookie only), so live 5h/weekly meters are not shown — behavior
  matches `opencode-go` and `9router`.

## 0.1.25 (2026-08-07)

### Improvements

- Widen peer dependency range to support Pi 0.84.0 (`>=0.80.8 <0.85.0`).
  No code changes — verified compatible against the 0.84.0 SDK types.

## 0.1.24 (2026-08-05)

### Improvements

- Patch version bump for release sync and package documentation update.

All notable changes to `pi-sub` will be documented in this file.

## 0.1.23 (2026-07-30)

### Improvements

- Widen Pi peer dependency range to <0.84.0 for Pi 0.83.0 compatibility.

## 0.1.22 (2026-07-30)

### Improvements

- Patch version bump for release sync and package documentation update.

## 0.1.21 (2026-07-24)

### Features

- Added 9router adapter for tokens per second (tok/s) and endpoint display.
- Cleared stale context on session shutdown to prevent crashes on `/new`.

## 0.1.15 (2026-07-16)

### Features

- Added `zai-coding-cn` (Z.ai China / open.bigmodel.cn) usage monitoring.
- Surfaced Z.ai MCP/month + per-model usage in `/sub` detail view.

## 0.1.10 (2026-07-10)

### Features

- Added tok/s (tokens per second) display for response speed tracking.

## 0.1.0 (2026-07-05)

### Features

- Initial release of `pi-sub` subscription usage footer extension.