Chat2API v1.4.0

🎉 Major Features

Managed Tool Calling Redesign

• OpenAI-compatible tool calling: Added a shared tool-calling engine for OpenAI-style tools, Cherry Studio MCP payloads, and managed tool prompts.
• Multiple protocol parsers: Added support for managed XML, managed bracket syntax, Anthropic-style tool use, and Codex Responses-style payloads.
• Streaming tool-call parsing: Added stream parsing for tool-call chunks so compatible clients can receive tool-call deltas more reliably.
• Runtime planning and diagnostics: Added provider profiles, tool-choice policy handling, runtime plan generation, and management API diagnostics for tool-calling behavior.

In-app Update Flow

• Updater manager integration: Update checks now go through the updater manager instead of ad hoc renderer logic.
• Renderer update status: Exposed updater status through preload and renderer APIs.
• Update UI states: The About page now reflects update checking, download, progress, install, and error states from updater events.
• Runtime app version display: The About page now reads the packaged app version instead of showing a hard-coded fallback.

Provider Model Refresh

• Website-aligned model catalogs: Refreshed default provider model lists from current web/API behavior.
• qwen-ai curated defaults: qwen-ai now keeps a curated default set while documenting the remaining website-supported models for manual addition.
• Provider documentation: Added per-provider docs to make future provider feature adaptation and user model-management workflows easier.

🚀 Enhancements

Provider Improvements

• GLM: Default model updated to GLM-5.1 with actual model id glm-5.1.
• Kimi: Removed stale Kimi-2.5 default and kept Kimi-2.6 aligned with the website.
• MiniMax: Removed stale MiniMax-M2.5 default and kept MiniMax-M2.7 aligned with the website.
• Qwen: Updated website-model defaults, including the Qwen3.6 display-name cleanup.
• qwen-ai: Updated default supported models from captured website/API behavior and documented additional official models for manual configuration.
• Z.ai: Refreshed default supported models from the latest captured web model list.
• DeepSeek: Improved web-mode support, search text handling, persisted model migration, and model mapping behavior.
• Perplexity: Updated documentation to list only the Free-user Auto model in the supported provider table.

Session and Proxy Behavior

• Kimi chat cleanup: Added provider-side session deletion support for Kimi.
• Qwen chat cleanup: Added provider-side session deletion support for Qwen.
• Provider dispatch cleanup: Simplified provider dispatch and adapter loading paths.
• Model mapping workflow: Improved model mapping form controls and synchronized provider selection when choosing mapped models.
• Proxy address display: Fixed the header proxy badge so it reflects the configured bind address instead of always showing 127.0.0.1.

Runtime and Maintenance

• Standalone app logs: Added standalone application log storage with dedicated manager/types and renderer store integration.
• Source artifact checks: Added build-time checks to prevent unwanted source artifacts from leaking into packaged output.
• Dashboard/log updates: Updated dashboard chart and log-store behavior to work with the new logging surface.

🐛 Bug Fixes

• Fixed DeepSeek web-mode compatibility and preserved search text behavior.
• Fixed persisted DeepSeek model migration for existing installations.
• Fixed built-in model mapping defaults and provider synchronization after selecting mapped models.
• Fixed update UI state wiring and repaired update page state syntax.
• Fixed About page version display so packaged builds show the actual application version.
• Fixed proxy header address display after changing the bind address.

📚 Documentation

• Added provider guides under docs/providers/ for DeepSeek, GLM, Kimi, Mimo, MiniMax, Perplexity, Qwen, qwen-ai, and Z.ai.
• Updated README and README_CN Supported Providers model tables.
• Added design and implementation notes for the in-app update flow.
• Added design and implementation notes for Chat2API tool-testing workflows.
• Documented tool configuration restore usage for local testing.

🧪 Testing and Tooling

• Added Chat2API testing skills for:
  • HAR tool fixture extraction
  • Management API checks
  • Provider model matrix runs
  • OpenAI-compatible client replay
  • Proxy behavior validation
• Added regression tests for:
  • Tool-calling engine, parser, stream parser, runtime plan, provider profiles, and tool-choice policy
  • Tool-calling management API and settings UI
  • DeepSeek stream handling and provider flow behavior
  • Header proxy address display
  • App log manager behavior
  • Source artifact checks
  • In-app update flow wiring

📦 Technical Details

Files Changed

• 124 files changed
• 10,617 lines added
• 1,649 lines removed

Release Artifacts

• macOS arm64/x64 DMG packages
• macOS x64 ZIP package
• Windows x64 installer and portable package
• Linux x64/arm64 AppImage packages
• Linux x64 tarball
• Debian amd64 package
• Auto-update metadata files for macOS, Windows, and Linux

🔄 Migration Notes

• Existing users should restart the app after upgrading so refreshed built-in provider model defaults can be synchronized into persisted provider settings.
• Users with custom qwen-ai models can keep using Provider Management / Model Management to add models that are supported by the website but not included in the curated default list.
• Provider-side chat cleanup is now available for Kimi and Qwen when Chat2API sessions are cleared.

✅ Verification

• npm run prebuild:check
• npm run build
• node --test tests/updater/in-app-update-flow.test.ts tests/providers/provider-flow.test.ts tests/tool-calling/*.test.ts
• Release workflow completed successfully for macOS arm64/x64, Windows x64, and Linux x64/arm64.

📝 Full Changelog

See v1.3.0...v1.4.0

---

Installation: Download the appropriate package for your platform from the assets below.

macOS Users: If you encounter "App is damaged and can't be opened", run:

sudo xattr -rd com.apple.quarantine "/Applications/Chat2API.app"