v5.4.9

Improvements

• Makes the installer default to a merged full catalog so new installs expose both modern base models and the full explicit preset catalog without hand-editing opencode.json.
• Keeps --modern and --legacy as explicit opt-outs while hardening installer writes with atomic temp-file persistence, Windows EPERM/EBUSY retry handling, and redacted error logging.
• Fixes --help to short-circuit before conflicting mode validation and preserves the full-catalog merge path with regression coverage.

Chores

• Validation: node --check scripts/install-opencode-codex-auth.js
• Validation: node scripts/install-opencode-codex-auth.js --modern --legacy --help
• Validation: npx vitest run test/install-opencode-codex-auth.test.ts
• Validation: git diff --check
• Validation: npm publish

Changelog

Full Changelog: v5.4.8...v5.4.9

• #100

v5.4.8

Improvements

• Adds JSON output support and routing visibility for read-only Codex ops so automation can inspect status, metrics, dashboard, and doctor flows without scraping text output.
• Adds a first-party ChatGPT device-code login flow for SSH, WSL, and other headless environments, and unifies login finalization across browser, manual, and device-code auth paths.
• Hardens runtime and account handling around storage import parity, deactivated-workspace rotation, and dependency-audit cleanup for the shipped package.

Chores

• Validation: npm ci
• Validation: npm run lint
• Validation: npm run typecheck
• Validation: npm run build
• Validation: npm test
• Validation: npm publish

Changelog

Full Changelog: v5.4.7...v5.4.8

• #92
• #93
• #94
• #95
• #96
• #97
• #99

v5.4.7

Improvements

• Corrects GPT-5 legacy alias routing so gpt-5-mini and gpt-5-nano normalize to the proper GPT-5.4 Mini/Nano families, adds first-class gpt-5.4-nano support, and tightens GPT-5.4 Pro reasoning coercion/validation.
• Refreshes shipped config metadata for the GPT-5.4 family and Codex models, including the current context-window values, new GPT-5.4 Pro/Nano entries, and updated config template guidance.
• Preserves caller-supplied max_output_tokens in transformed Responses requests instead of silently clearing explicit output budgets.
• Treats built-in Responses tools as first-class bridge/runtime entries and expands passthrough coverage for newer Responses API request fields.
• Adds codex-doctor prompt-cache diagnostics with masked telemetry, safer abort handling, and stronger request-level prompt-cache accounting.

Chores

• Validation: npm ci
• Validation: npm run typecheck
• Validation: npm test
• Validation: npm publish

Changelog

Full Changelog: v5.4.6...v5.4.7

• #87
• #88
• #89
• #90
• #91

v5.4.6

Improvements

• Gracefully handles deactivated_workspace responses by removing and flagging only the dead workspace entry instead of dropping healthy siblings that share the same user or refresh token.
• Preserves workspace identity during doctor/health cleanup and keeps deactivated workspaces out of verify-flagged restore and quota-probe paths so the wrong account is not restored or removed.
• Expands regression coverage for deactivated-workspace detection, org-scoped cleanup, and flagged-account restore behavior.

Chores

• Validation: npm run lint
• Validation: npm run typecheck
• Validation: npm run build
• Validation: npm test
• Validation: npm pack --dry-run

Changelog

Full Changelog: v5.4.5...v5.4.6

• #85

v5.4.5

Improvements

• Added first-class gpt-5.4-mini support across model normalization, prompt-family routing, shipped config templates, and configuration docs.
• Expanded regression coverage for GPT-5.4 Mini model mapping, prompt-cache isolation, request transformation, and reasoning behavior while preserving legacy gpt-5-mini / gpt-5-nano alias semantics.
• Refreshed the README/docs presentation and repository metadata surfaces for the current operational workflow.
• Added anti-slop PR screening guardrails via the repository PR template and quality workflow.

Chores

• Validation: npm run lint
• Validation: npm run typecheck
• Validation: npm run build
• Validation: npm test

Changelog

Full Changelog: v5.4.4...v5.4.5

• #81
• #83

v5.4.4

Improvements

• Hardened codex-limits usage fetching with AbortController timeouts, capped/redacted upstream error handling, and consistent abort-path coverage.
• Fixed credential deduplication so accounts without refresh tokens are preserved instead of collapsing into a single entry.
• Preserved active-account labeling across deduplicated credentials and aligned fetched usage context with the displayed active account.
• Moved refreshed-credential persistence onto fresh storage transactions to avoid stale-write races and improve Windows file-lock resilience.
• Tightened fallback identity matching so same-email accounts from different workspaces cannot be overwritten during refresh recovery.
• Expanded regression coverage for refresh fan-out, transactional fallback identity handling, blank-token refresh rejection, redaction behavior, and multiple abort/read failure paths.

Chores

• Validation: npm run typecheck
• Validation: npm run lint
• Validation: npm test
• Validation: npm run build

Changelog

Full Changelog: v5.4.3...v5.4.4

• #72

v5.4.3

Improvements

• Added GPT-5.4 snapshot alias normalization for gpt-5.4-2026-03-05* and gpt-5.4-pro-2026-03-05* IDs (including effort suffix variants).
• Updated legacy GPT-5 aliases (gpt-5, gpt-5-mini, gpt-5-nano) to normalize to gpt-5.4 as the default general family.
• Split GPT-5.4 Pro prompt-family handling from GPT-5.4 for independent prompt/cache routing while preserving fallback behavior (gpt-5.4-pro -> gpt-5.4).
• Updated OpenCode templates/docs for GPT-5.4 1M context defaults and optional compact-window tuning guidance.
• Expanded regression coverage for snapshot aliases, pro-family separation, and legacy alias normalization.

Chores

• Validation: npm run typecheck
• Validation: npm run lint
• Validation: npm test
• Validation: npm run build

Changelog

Full Changelog: v5.4.2...v5.4.3

• #67

v5.4.2

Improvements

• Added GPT-5.4 runtime support as the latest general-purpose family across model normalization/map, prompt-family routing, quota probes, and fallback handling.
• Added optional/manual GPT-5.4 Pro support, including the default fallback edge gpt-5.4-pro -> gpt-5.4 in fallback mode.
• Hardened GPT-5.4 family detection with boundary-aware matching to prevent accidental normalization/family routing for nearby version strings (for example gpt-5.4).
• Refreshed config templates and documentation to default to GPT-5.4 and document GPT-5.4 Pro reasoning/coercion behavior.
• Expanded regression coverage for GPT-5.4/GPT-5.4-pro normalization, reasoning, fallback, and edge-case routing.

Chores

• Validation: npm run typecheck
• Validation: npm run lint
• Validation: npm test
• Validation: npm run build

Changelog

Full Changelog: v5.4.1...v5.4.2

• #65

v5.4.1

Improvements

• Hardened auth-refresh failure handling for grouped removals by guarding the zero-removal path and applying refresh-token-scoped cooldown to live accounts.
• Cleared stale in-memory auth-failure counters when refresh tokens rotate during auth updates to prevent false-positive removals.
• Added regression coverage for zero grouped-removal handling and rotated refresh-token failure-state cleanup.

Chores

• Validation: npm run typecheck
• Validation: npm run lint
• Validation: npm test
• Validation: npm run build

Changelog

Full Changelog: v5.4.0...v5.4.1

• #64

v5.4.0

Improvements

• Preserved distinct org-scoped workspace identities when organizationId is shared but accountId differs, across authorize-time matching and refresh-token collision pruning.
• Aligned no-org duplicate handling to collapse only true identity-compatible records while keeping active-index remapping stable.
• Hardened id-token organization binding to canonical idToken['https://api.openai.com/auth'].organizations[0].id sourcing.

Chores

• Validation: npm run lint
• Validation: npm run typecheck
• Validation: npm test
• Validation: npm run build

Changelog

Full Changelog: v5.3.4...v5.4.0

• #41