@kikkimo/claude-launcher 2.5.0 → 3.1.0

package/CHANGELOG.md

@@ -5,6 +5,76 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [3.1.0] - 2026-05-09
+
+### Added
+- **6-Step Add API Wizard**: Multi-step wizard (Provider → URL → Token → Model → Name → Confirm) with state machine. Esc/back navigation between steps, pre-create duplicate detection with branch UI, draft editing before persistence, and race-condition handling at persist time.
+- **Draft Layer**: `buildApiDraft()` / `applyDraftEnvChange()` / `deleteDraftCustomEnvVar()` static methods on `ApiManager` for pre-persist config editing without touching disk.
+- **Model Tier Auto-Matching**: Same-generation tier templates for all 9 providers — Anthropic auto-detects Opus/Sonnet/Haiku from model list; DeepSeek maps pro→flash; GLM maps 5.1→5-turbo; Moonshot maps k2.6→k2-thinking-turbo; MiniMax maps M2.x→M2.x-highspeed. Subagent defaults to Haiku tier (per Anthropic recommendation).
+- **Provider Default Values**: All 9 providers now carry runtime env defaults (`API_TIMEOUT_MS`, `CLAUDE_CODE_ATTRIBUTION_HEADER: '0'`, `CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS: '1'`, etc.). DeepSeek defaults `CLAUDE_CODE_EFFORT_LEVEL` to `max`. Moonshot/Kimi/MiniMax/GLM default `CLAUDE_CODE_DISABLE_NONSTREAMING_FALLBACK: '1'`.
+- **Comprehensive Hint System**: 12 field-level `_detail` hints across all 11 locales — each shows field description + corresponding env var `[KEY]` + dynamic value source (provider default / manual / built-in default). Both list pages and edit sub-pages covered.
+- **New i18n Sections**: `page`, `action`, `prompt`, `add_api`, `summary`, `confirm`, `config_labels.model/runtime` — ~50 new keys across all 11 languages.
+- **New i18n Keys**: `hints.model.*_detail`, `hints.runtime.*_detail`, `hints.runtime.source_*`, `hints.runtime.effort_values`, `action.cancel_config`, `add_api.confirm_page_prompt`, `errors.api.not_found`, `status.auto`, `config.values.recommended_on`, `navigation.enter_to_edit/select`, `navigation.input_1_to_n_or_q`, `navigation.invalid_selection`.
+- **Menu Component** `navigationKey` param + `_navigationKey` cache passed through all redraws. Non-TTY fallback now renders numbered prefixes (`1.`, `2.`) and i18n prompts. `selectFromList()` prompts i18n'd.
+
+### Changed
+- **Runtime Config Display**: "Default" replaced with actual provider values (e.g. `600000`, `0`, `max`). TYPE_A `'1'` shown as "Enabled", `'off'` as "Disabled". Fields without provider default shown as `(not set)` via new `status.auto` key.
+- **Config Label i18n**: `i18nLabel()` helper resolves `config_labels.<section>.<key>` lookup before falling back to English constants. Labels now follow user locale.
+- **Edit API Menu**: 3 env entries merged into single "Model & Runtime Config" entry with summary counts (7→5 items). Sub-pages restructured as 3-section home with per-section hints.
+- **Model/Runtime List Pages**: Dynamic column alignment via `getStringWidth()`/`padStringToWidth()` instead of hardcoded `padEnd(24)`. Labels align correctly in all languages.
+- **Experimental Features Label**: All 11 locales updated from ambiguous "Experimental Features" to "Disable Experimental Features" matching `CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS` semantics.
+- **noFlicker Default**: Now defaults to `true` (On, recommended) matching telemetry's default pattern. Added `config.values.recommended_on` key.
+- **API Default Name**: Simplified from `"Provider Name model-name"` to `"ProviderShort #N"` with auto-increment counting by short-name prefix (moonshot + kimi_for_coding share "Moonshot AI" prefix).
+
+### Fixed
+- **Confirmation Page Rendering**: Content passed via `versionInfo` parameter to Menu instead of `screen.render()` that was immediately overwritten by `displayMenu()`.
+- **Confirmation Page Back Button**: Changed from "Back" to "Cancel" (`action.cancel_config`) — exits to main menu without saving.
+- **Token Input**: Removed misleading "empty to restore recommended" hint; added minimum 10-character validation.
+- **Exit Handling**: `addNewThirdPartyApi` state machine wrapped in try-catch so `exit` at any step returns to main menu gracefully.
+- **Locale File Structure**: Fixed `nonstreaming` key having `},` on the same line, premature `runtime` section closure causing `source_*` keys to land at wrong nesting level.
+
+## [3.0.0] - 2026-04-07
+
+### Added
+- **ANSI Screen Rendering Layer**: Full terminal rendering rewrite using alternate screen buffer (`\x1b[?1049h`) and absolute cursor positioning (`\x1b[H\x1b[2J`). Eliminates all position drift across page transitions. Program runs in isolated screen like vim/htop; exit restores original terminal content.
+  - New `lib/ui/screen.js` singleton: `render()`, `write()`, `enter()`, `exit()`, `exitForHandoff()`, `debug()`, `showCursor()`/`hideCursor()`, `setReadlineActive()`, `isActive()`
+  - Test mode (`SCREEN_TEST=1`): source-tagged write interception for automated leak detection
+  - Degradation: non-TTY passthrough, `SCREEN_NO_ALT=1` manual override
+- **Edit API Feature**: New menu item to modify API name, provider, base URL, and model (API key not editable)
+  - Field-by-field editing with per-field validation (reuses Add API validators)
+  - Provider selection via preset list (not free text), preserves provider-specific envVars/timeout/upgrade detection
+  - Provider/URL mismatch warning in field menu hint area
+  - Auto-save per field edit with success/cancel feedback
+- **Unified Password Guard**: Shared `passwordGuard()` function protects delete, edit, import, export operations
+  - Mode A (delete/edit): guard at dispatch layer with header display
+  - Mode B (export/import): guard inside handler after title page
+  - Handles wrong password, empty password, Esc cancel, Ctrl+C delegation distinctly
+  - Defense-in-depth: export/import return false when no password set
+- **API Table Pagination**: ←→ page navigation for API selection tables (remove/switch/edit)
+  - Dynamic items-per-page based on terminal height
+  - Per-page selection memory across page switches
+  - 3 pure testable helpers: `calculatePagination()`, `initPaginationState()`, `handlePageKeyPress()`
+  - Legacy >99 API defensive guard with display truncation + warning
+- **API Count Limit**: Maximum 99 APIs enforced in `addApi()` and import path
+- **Launch Handoff Lifecycle**: Clean `screen.exitForHandoff()` → normal terminal output → `relinquishConsoleToChild()` → spawn sequence
+  - `handleLaunchFailure()` promoted to module-level with `rollbackFn(errorMessage)` callback chain
+  - Pre-handoff errors show in alt-screen; post-handoff errors use press-key + 60s timeout + exit
+- **Menu Hint Enhancements**: Password-required hints (🔒) for edit/remove/export/import when password is set
+- **Navigation i18n**: Action words (edit/remove/switch/select) fully localized across 11 locales
+
+### Changed
+- **All Terminal Output**: 462 direct `console.clear/log/error/warn` and `process.stdout.write` calls replaced with `screen.render()`/`screen.write()`/`screen.debug()` across 12 files
+- **Menu Component**: `displayMenu()` and `navigate()` now use `screen.render()` for absolute positioning; `clearScreen` parameter removed from `navigate()` signature
+- **Interactive Table**: Refactored to `screen.render()` with pagination support; action text localized via i18n keys
+- **Signal Ownership**: Global SIGINT handler respects `handleCtrlC()` return value (first Ctrl+C = warning only); SIGTERM/uncaughtException/unhandledRejection handlers call `screen.exit()` before exit
+- **Launcher Lifecycle**: `relinquishConsoleToChild()` moved before `spawn()` for clean handoff; `updateApiModel()` delegates to `updateApiField()` for unified validation
+- **Default Config Language**: Changed from `zh` to `en` in `loadConfig()`/`loadConfigSync()` to match `LanguageManager` default; config file written on first run
+- **Hint Area Spacing**: Extra space after ℹ icon; multi-line hint indentation aligned
+
+### Fixed
+- **First-run Language Bug**: Deleting config and restarting no longer switches from English to Chinese on second launch
+- **Field Menu CJK Alignment**: Label padding uses `getStringWidth()`/`padStringToWidth()` for correct CJK character width
+
 ## [2.5.0] - 2026-03-31
 
 ### Added

package/README.md

@@ -13,24 +13,32 @@ An elegant interactive launcher for Claude Code with a beautiful Claude-style in
 
 ### 🎨 **Beautiful Interface**
 - Claude-style interface with authentic orange/amber color scheme
+- ANSI alternate screen buffer for drift-free rendering (like vim/htop)
 - Arrow key navigation with smooth menu transitions
+- Paginated API tables with ←→ page navigation for large API lists
 - Interactive tables for API selection and management
 - Multi-language support (English, Simplified Chinese, Traditional Chinese, German, French, Spanish, Italian, Portuguese, Japanese, Korean, Russian)
 
 ### 🔐 **Advanced Security**
 - AES-256-CBC encryption for all sensitive data
 - Machine-specific encryption keys for enhanced security
+- Unified password guard for high-risk operations (edit, delete, import, export)
 - Password-protected configuration import/export
 - Secure API token storage with masked display
 - Strong password requirements and validation
 
 ### 🚀 **Third-party API Management**
-- Full support for multiple third-party API providers (Anthropic, DeepSeek, Kimi K2.5, MiniMax M2.7, GLM-5.1/ZhiPu AI, and custom APIs)
-- Interactive API configuration with validation
+- Full support for 9 third-party API providers (Anthropic, DeepSeek, Kimi K2.6, MiniMax M2.7, GLM-5.1/ZhiPu AI, and custom APIs)
+- **6-Step Add API Wizard**: Provider → URL → Token → Model → Name → Config Confirm with pre-create duplicate detection, back navigation between steps, and inline config editing before persist
+- **Env Config Editor**: Edit model config (6 fields: Sonnet/Opus/Haiku/Subagent/Custom), runtime config (6 fields: timeout/attribution/nonessential/effort/experimental/nonstreaming), and custom env vars — with per-field hints, provider default values, and overridden markers
+- **Auto Model Tier Matching**: Same-generation auto-matching for all providers (AnthropicOpus/Sonnet/Haiku, DeepSeek pro/flash, GLM 5.1/turbo, etc.)
+- **Interactive API Editing**: Modify name, provider, base URL, model, and all env configs for existing APIs
+- Dynamic column alignment and comprehensive field hints across all supported languages
 - API usage statistics with success/failure tracking
 - Model upgrade notifications and auto-upgrade support
 - Secure configuration backup and restore
 - Easy API switching, removal, and bulk clear
+- Maximum 99 APIs supported per configuration
 
 ### 🌍 **Enterprise-grade Features**
 - Global installation - use `claude-launcher` from anywhere
@@ -88,20 +96,21 @@ node claude-launcher
 4. **Launch Claude Code with Third-party API** - Use configured third-party API
 5. **Launch Claude Code with Third-party API (Skip Permissions)** - Combine third-party API with permission skipping
 6. **Third-party API Management** - Full API lifecycle management:
-   - Add, switch, and remove APIs
+   - Add, edit, switch, and remove APIs
    - View usage statistics with success/failure rates
    - Model upgrade settings (auto/manual upgrade)
-   - Import/export configurations
-7. **Language Settings** - Switch between 11 supported languages
+   - Import/export configurations (password-protected)
+7. **Configuration Management** - Language, telemetry, launch mode, model upgrade settings
 8. **Version Update Check** - Check for launcher updates
 9. **Exit** - Close the launcher
 
 ### Interactive Navigation
 
-- **Arrow Keys**: Use ↑↓ to navigate, Enter to select
-- **Escape Key**: Press Esc to go back or exit
+- **Arrow Keys**: Use ↑↓ to navigate, ←→ to switch pages (in paginated tables), Enter to select
+- **Escape Key**: Press Esc to go back or cancel
+- **Ctrl+C**: First press shows warning, second press exits cleanly
 - **Multi-language**: All interface text adapts to your selected language
-- **Smart Tables**: Interactive tables for API management with clear visual feedback
+- **Smart Tables**: Paginated interactive tables for API management with per-page selection memory
 
 ### Example Session
 
@@ -120,7 +129,7 @@ $ claude-launcher
     Launch Claude Code with Third-party API
     Launch Claude Code with Third-party API (Skip Permissions)
     Third-party API Management
-    Language Settings
+    Configuration Management
     Version Update Check
     Exit
 ```
@@ -133,12 +142,13 @@ Access comprehensive API management through the dedicated menu:
 📋 Third-party API Management
 
   → Add New API
+    Edit API            → Select API → Edit name/provider/URL/model
     Remove API          → Delete Single API / Clear All APIs
     Switch Active API
     View Statistics     → View Details / Reset Statistics
     Model Upgrade       → Auto Upgrade [ON/OFF] / Manual Upgrade
-    Export Configuration
-    Import Configuration
+    Export Configuration  🔒 (password required)
+    Import Configuration  🔒 (password required)
     Change Password
     Back to Main Menu
 ```
@@ -154,7 +164,7 @@ The launcher automatically checks for model upgrades when you start:
 
 ### Modern Configuration System
 
-Claude Launcher 2.0 uses an advanced configuration system:
+Claude Launcher uses an advanced configuration system:
 
 1. **Encrypted JSON Storage**: Configuration stored at `~/.claude-launcher-apis.json`
 2. **Interactive Setup**: First-time wizard guides you through all options
@@ -173,7 +183,7 @@ Claude Launcher 2.0 uses an advanced configuration system:
 
 Configure any third-party API provider through the interactive interface:
 
-- **Supported Providers**: Anthropic, DeepSeek, Moonshot/Kimi (K2.5), MiniMax (CN/Global, M2.7), GLM/ZhiPu AI (GLM-5.1), and custom Anthropic-compatible APIs
+- **Supported Providers**: Anthropic (Opus 4.7/Sonnet 4.6/Haiku 4.5), DeepSeek (V4-Pro/V4-Flash), Moonshot/Kimi (K2.6), MiniMax CN/Global (M2.7), ZhiPu AI/Z.ai (GLM-5.1/5-Turbo), and custom Anthropic-compatible APIs
 - **Secure Storage**: All API tokens encrypted before storage
 - **Validation**: Real-time validation of URLs, tokens, and models
 - **Usage Tracking**: Monitor API usage statistics with success/failure rates