- enable Fast for Codex API, OpenAI-compatible, and Codex OAuth sources
- apply service_tier=priority as an independent payload override
- support Fast-only and reasoning-effort combinations
- disable Fast until a supported model source is selected
- Decouple reasoning effort from the Fast service tier
- Generate native override parameters for each provider
- Validate efforts against the model’s supported levels
- Preserve complete reasoning capabilities on configured models
- Add coverage for multi-provider alias overrides
* chore(release): stop publishing legacy Windows update packages
* feat(agents): support OpenCode Desktop across platforms
Detect CLI and desktop installations on Windows, macOS, and Linux, and add separate launch and restart controls.
Closes#156
* fix(codex): align model display names and show total session count
- preserve runtime model identity in generated catalogs
- count distinct sessions across database and rollout sources
- update session count labels in all supported languages
* fix(codex): keep config and auth files in sync
- generate auth.json for API key configuration
- preserve and validate OAuth credentials
- include auth.json in transactional updates and rollback
- retain compatibility with legacy Codex configuration states
* fix(cpa-gui): update version to 0.2.59 in Cargo files
* fix(codex): preserve OAuth mode after closing configuration
* feat(config): add TLS management and protocol-aware local URLs
- add TLS controls to advanced settings
- persist certificate and private key paths
- restart the core after TLS changes
- switch WebUI, API and management URLs between HTTP and HTTPS
- synchronize agent configurations and detection with the active protocol
* refactor(config): reorganize network, routing, and TLS settings
- split network settings and session routing into separate tabs
- move TLS configuration into network settings
- preserve TLS restart and HTTPS URL synchronization behavior
* feat(config): add independent network settings and custom listen IP
* fix(navigation): allow access to usage records while core is stopped
* refactor(home): show a single API URL based on the configured listen IP
* feat(config): add file pickers for TLS certificate and private key
* fix(usage): distinguish client cancellations from request failures
- persist usage failure status and response details
- classify HTTP 499 and canceled contexts separately
- exclude canceled requests from failure statistics
- add canceled filtering and failure details to the usage UI
- migrate and backfill existing usage records
* feat(ui): revamp usage records and improve app interactions
- redesign usage records dashboard, filters, charts, and pricing views
- add resizable centered columns with persisted widths
- expose OpenCode desktop launch and restart actions
- suppress the API access enabled success notice
* feat(agents): persist applied configurations across app restarts
- preserve applied agent configs and managed Claude aliases on exit
- keep explicit close actions restoring original configurations
- add Codex restart persistence regression coverage
- bump desktop version to 0.2.61
* fix(model-aliases): support provider-aware OAuth aliases
- map OAuth aliases to their correct provider channels
- support Antigravity aliases with force mapping
- read and delete aliases by their exact OAuth channel
- refresh the runtime model registry after alias changes
- restrict Fast aliases to GPT models from OpenAI-compatible or Codex sources
* docs(readme): add APIMart sponsorship sections
- add English and Chinese sponsor descriptions
- add localized APIMart banner images
- support creating plain model aliases without requiring Fast
- enable the Fast toggle before an original model is selected
- preserve Fast state for compatible models and handle plain alias deletion
lzt404