Add Gemini, Qwen, and OpenRouter sessions - #55
Conversation
|
👋 Hello @glenn-jocher, thank you for submitting a
For more guidance, please refer to our Contributing Guide. Don't hesitate to leave a comment if you have any questions. Thank you for contributing to Ultralytics! 🚀 |
UltralyticsAssistant
left a comment
There was a problem hiding this comment.
🔍 PR Review
Made with ❤️ by Ultralytics Actions
Reviewed all 9 changed files, including the Rust launch/auth paths, provider registry, session dialog, settings status, labels, process detection, and documentation. The provider metadata refactor is coherent, but the Gemini launch path is currently unusable, Lite-owned Qwen keys are not passed under Qwen's OpenAI auth mode, and Qwen auth status can be falsely reported as configured. Remaining risk is concentrated in native CLI compatibility and provider-auth state detection.
💬 Posted 3 inline comments
⚠️ HIGHsrc-tauri/src/lib.rs:1996Gemini CLI's current command-line reference supports--resumebut does not define--session-id; new Gemini sessions therefore launch with an unsupported option and fail before opening the harness. Removing this flag alone would also leave Lite's UUID unrelated to Gemini's native session ID, so resumability would still be broken. Use a supported new-session path and capture/pass Gemini's native session ID for later--resume.⚠️ HIGHsrc-tauri/src/lib.rs:2018When Lite owns a Qwen key, this selects Qwen'sopenaiauth mode, but the later environment injection still sets onlyDASHSCOPE_API_KEY. Qwen's OpenAI-compatible mode resolvesOPENAI_API_KEYby default unless anenvKeymodel-provider entry exists, and this launch path creates no such entry; the saved Lite key is therefore not supplied to Qwen. Pass the key throughOPENAI_API_KEYfor this branch or use Qwen's--openai-api-keyargument. ([qwenlm.github.io](https://qwenlm.github.io/qwen…- 📝 LOW
src-tauri/src/lib.rs:1708The existence of~/.qwen/settings.jsonis not evidence that Qwen is authenticated: it is the general user settings file and may contain only preferences. This causes Settings to hide the Sign in action and displays Qwen as configured for users who have no auth provider or credentials configured. Check an auth configuration marker such as the selected auth/provider settings rather than file presence alone. ([qwenlm.github.io](https://qwenlm.github.io/qwen-code-docs/en/users/configuration/setti…
|
🎉 PR #55 has been merged—thank you, @glenn-jocher, for this excellent contribution to Lite!
Lite 0.0.14 now supports Gemini CLI, Qwen Code, and OpenRouter, while streamlining CLI setup and making session close and restart actions reversible during the eight-second Undo window. Great work! |
Summary
0.0.14with Gemini CLI and Qwen Code as resumable harnesses using their native sign-in flows and Portal-owned Google/Alibaba iconsProvider behavior
GEMINI_API_KEYor its native Google sign-inOPENROUTER_API_KEYand~openai/gpt-latestValidation
bun run checkbun run buildcargo fmt --check --manifest-path src-tauri/Cargo.tomlcargo check --manifest-path src-tauri/Cargo.tomlgit diff --check🛠️ PR Summary
Made with ❤️ by Ultralytics Actions
🌟 Summary
Lite 0.0.14 adds Gemini CLI, Qwen Code, and OpenRouter support while centralizing provider setup, CLI installation, and deferred session actions.
📊 Key Changes
OPENROUTER_API_KEYand~openai/gpt-latest, with DeepSeek and OpenRouter sharing generic Codex-provider handling.🎯 Purpose & Impact
📋 Skipped 1 file (lock files, generated, images, etc.)
src-tauri/Cargo.lock