Skip to content
Permalink

Comparing changes

Choose two branches to see what’s changed or to start a new pull request. If you need to, you can also or learn more about diff comparisons.

Open a pull request

Create a new pull request by comparing changes across two branches. If you need to, you can also . Learn more about diff comparisons here.
base repository: ClaudeCodeUsage/ClaudeCodeUsage
Failed to load repositories. Confirm that selected base ref is valid, then try again.
Loading
base: v2.2.1
Choose a base ref
...
head repository: ClaudeCodeUsage/ClaudeCodeUsage
Failed to load repositories. Confirm that selected head ref is valid, then try again.
Loading
compare: v2.2.2
Choose a head ref
  • 4 commits
  • 15 files changed
  • 4 contributors

Commits on Aug 1, 2026

  1. fix: track Claude Opus 5 as a 1M-context model (#81) 

    Claude Code writes the plain `claude-opus-5` id to its JSONL even when the
session runs the 1M-context variant, so the existing `[1m]` marker never
fired for it. `contextWindowFor` fell through to the generic opus/sonnet
branch and returned 200K flagged as exact, making the reported context
percentage five times too high with no `estimated` hint to soften it.

The major-version branch now covers Opus 5+ as well as Sonnet 5+. Pre-4.6
Opus/Sonnet and dated 3.x snapshots keep their 200K window (the alternation
still cannot match a date suffix such as `sonnet-20241022`).

`claude-opus-5` also gets an explicit pricing entry. The family fallback
already resolved it to the correct $5/$25 current-Opus tier, so no cost
figure changes; the entry mirrors the existing `claude-sonnet-5` anchor and
drops the spurious `Unknown model` line from the diagnostic output.

Opus 5 pricing and context window verified 2026-07-28 against
https://platform.claude.com/docs/en/about-claude/models/overview
    @e7d
    e7d authored Aug 1, 2026
    Configuration menu
    Copy the full SHA
    f5c2e10 View commit details
    Browse the repository at this point in the history

  2. fix: one unprobeable JSONL no longer blanks usage (#79) 

    sortUsageFilesByEarliestTimestamp mapped readEarliestTimestamp over every
file with no per-file error handling. A file that trips the probe's 1 MiB
no-timestamp guard (TimestampProbeLimitError), such as a large workflow
journal.jsonl, rejected the whole batch, so loadUsageRecords returned empty
records and the dashboard blanked on every refresh (failed=1, bytes=0).

Catch the per-file failure and fall back to timestampMs 0 so the file sorts
by discovery order, matching the existing small-no-timestamp behavior. The
rest of the load proceeds and the file still reaches the main reader.
readEarliestTimestamp and its test are unchanged. Adds a regression test.

Co-authored-by: Carl <74130028+Carl723000@users.noreply.github.com>
Co-authored-by: OpenAI Codex <215057067+openai-codex[bot]@users.noreply.github.com>
    @ptweezy @Carl723000 @openai-codex
    3 people authored Aug 1, 2026
    Configuration menu
    Copy the full SHA
    231a30b View commit details
    Browse the repository at this point in the history

  3. fix: reduce background refresh energy use for v2.2.2 (#83) 

    Suspend polling and file watchers in unfocused VS Code windows, refresh on focus recovery, and back off repeated quota authentication failures.\n\nCo-authored-by: OpenAI Codex <215057067+openai-codex[bot]@users.noreply.github.com>
    @Carl723000
    Carl723000 authored Aug 1, 2026
    Configuration menu
    Copy the full SHA
    37d3255 View commit details
    Browse the repository at this point in the history

  4. docs: finalize v2.2.2 release notes (#85) 

    Move the post-v2.2.1 contributor fixes into the v2.2.2 section while preserving their PR attribution.

Co-authored-by: OpenAI Codex <215057067+openai-codex[bot]@users.noreply.github.com>
    @Carl723000 @openai-codex
    Carl723000 and openai-codex[bot] authored Aug 1, 2026
    Configuration menu
    Copy the full SHA
    cbee90e View commit details
    Browse the repository at this point in the history
Loading