feat: release 2.5.0 achievements and compressed Codex history by Carl723000 · Pull Request #131 · ClaudeCodeUsage/ClaudeCodeUsage

Carl723000 · 2026-10-14T12:11:26Z

Summary

Release 2.5.0 combines the local achievement system with compressed Codex history support and bounded sharing/reliability improvements.

Local achievement gallery: 50 themes and 178 conditions, Bronze/Silver/Gold labels, retained-log unlock dates, manual recalibration, expand/collapse controls, compact colored rank rows, and current totals on fully earned cards.
Claude and Codex weekly API-equivalent achievements add Platinum ($10,000) and Legendary ($25,000), preserving existing IDs, awards and frozen reference-price provenance.
Read .jsonl.zst Codex session/archive logs locally without rewriting sources or creating plaintext disk copies. Canonical source selection, lineage deduplication, verified EOF/fingerprints and incremental numeric caches prevent duplicate totals and stale evidence promotion. Unsupported hosts or failed reads retain last-good results with an explicit status.
Share & compare navigation, provider/metric heatmaps, improved badge selection/resource loading, and a public repository link in achievement SVG exports.
Optional, default-off weekly quota pace; bounded evidence recovery and large Claude-history argument-stack overflow prevention (reported by @mkgaskin-ops in RangeError: Maximum call stack size exceeded in recomputeDerivedRows once one project exceeds ~125k usage records[Bug]: #126).

Estimates remain distinct from bills and subscription balances; missing evidence is not zero. Independent root tasks do not imply completed work. Durable usage archives, broader sharing-page visual unification, and reliable multi-account attribution remain deferred. No code from the still-open #127, #128 or #129 is represented as shipped here.

Validation and acceptance

Positive candidate-test feedback was recorded on 2026-10-15.
Accepted runtime candidate: f73b14fb9c4f5e899ba69500d59fce40f357ce70. Subsequent release-readiness documentation changes do not change production runtime or assets.
Release head: 8bc07ef97bc25516faaeb2892ea48a7801ef2c8e. Full CI exposed one stale Compare accessible-name expectation after the accepted Share & compare rename. It was reproduced in the canonical isolated Linux container and corrected without changing production code; all three focused provider-name checks pass. Fresh full CI: https://github.com/ClaudeCodeUsage/ClaudeCodeUsage/actions/runs/38746617291 .
Local Node 24 suite: 2,647 passed / 2 skipped. Node 20: 2,621 passed / 28 skipped; native-codec availability accounts for the difference.
Node 22 compressed-log focused validation: 70/70. Latest focused canonical Linux browser run: 127/127, with no retries or page errors; this is not a claim of a complete local browser suite or agent-performed native/F5 acceptance.
Full fresh GitHub CI passed at the exact release head: Node Tests, all 876 browser checks, UI Test aggregate and Release Package. The squash commit d6d69aa246a277bdbf5b1a987b9ead71d8a799e2 has one parent and its tree exactly matches that tested head.
Candidate package identity, safety inventory and compiled runtime/assets were checked. Source manifests intentionally retain their development version; the publish workflow stamps 2.5.0 from the reviewed GitHub Release tag.

Integration and release

Use Squash and merge, preserving @Carl723000 as author and the applicable OpenAI Codex coauthor trailer. Apply minor so Release Drafter prepares v2.5.0. Publish the reviewed existing draft at the verified squash commit, then verify the GitHub VSIX, VS Code Marketplace and Open VSX delivery. No manual manifest bump or manual tag creation.

Squash merged on 2026-10-15. v2.5.0 is published at the verified squash SHA. Automated delivery: https://github.com/ClaudeCodeUsage/ClaudeCodeUsage/actions/runs/38750266268 .

🤖 Generated with OpenAI Codex

Selectively port evidence-backed badges and sharing, preserve source-scoped caches, add a once-only upgrade disable action, revoke stale award proof, and keep historical awards quiet. Co-authored-by: OpenAI Codex <215057067+openai-codex[bot]@users.noreply.github.com>

Use one-pass time bounds and bounded appends without changing Token accounting; cover 150k-record groups and measure installed-resource renderer paths. Co-authored-by: OpenAI Codex <215057067+openai-codex[bot]@users.noreply.github.com>

Align all nine README editions and both architecture copies with default-on opt-out, quiet historical unlocks, bounded hour proof, and preview-first badge sharing. Co-authored-by: OpenAI Codex <215057067+openai-codex[bot]@users.noreply.github.com>

Cover the peak-memory reporting invariant and label retained synthetic snapshots separately from a production soak. Co-authored-by: OpenAI Codex <215057067+openai-codex[bot]@users.noreply.github.com>

Document bounded rule, action, hour and weekly-peak storage in both user-facing inventories and enforce their shared exact fields. Co-authored-by: OpenAI Codex <215057067+openai-codex[bot]@users.noreply.github.com>