stacklok / toolhive-studio / 25870745122 / 1

70%

main: 70%

push

github

feat(playground): show per-message cost using models.dev pricing (#2255)

* feat(playground): show per-message cost using models.dev pricing

Adds a per-message USD cost next to the existing token breakdown on
assistant messages. Cost is computed from models.dev pricing
(USD/1M tokens), fetched once and cached on disk for 24h in the main
process. Cached input tokens use the model's cache_read rate when
available. Providers/models with no pricing entry (local providers,
unknown models) render exactly as before.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* chore(chat): drop unused ModelCost export from main pricing module

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(playground): address Copilot review feedback on chat cost

- useModelPricing now only enables its React Query fetch when a remote
  provider+model pair is known (skips ollama/lmstudio and the empty
  state). Avoids a no-op IPC + models.dev fetch for local-only users.
- Inline cost rendering no longer strips the dollar sign from the
  formatter output. The "<$0.01" case now reads correctly inline
  instead of "<0.01".
- Add unit tests for the main-process pricing module: cold-start
  fetch + write, warm disk cache within TTL, stale cache returns
  immediately + background refresh, offline with no disk cache, and
  non-OK upstream response.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

5107 of 7819 branches covered (65.32%)

7597 of 10747 relevant lines covered (70.69%)

116.48 hits per line