Desktop v0.0.27

• An expired Cline sign-in now produces one actionable error instead of two dead ends. A turn that fails before the runtime takes the prompt was reported twice — the hub's detail-less run.failed appended a bubble, then the send RPC resolved with the real error and overwrote the error banner underneath it — so you got a vague bubble stacked on a detailed banner. The failure now renders exactly once, upgrading in place when the detailed report arrives. Credential failures also carry a fix button: Sign in to Cline for the Cline provider (Settings → API Providers reads a stale token as "Signed in via browser", so there is nothing to fix there), Open model settings for others, and local-CLI providers keep pointing at their CLI. Sessions that fail to start over credentials get the same hint and action
• The Account page now offers a sign-in prompt when your Cline refresh token is rejected, instead of an error card whose Retry failed the same way. A rejected refresh was falling back to the persisted access token, which is dead too, so the account request 401'd; a rejected refresh now reports as signed out. Transient refresh failures still fall back
• Signing out of Cline and Cline Pass now sticks. Sign-out removes the provider entry, but the runtime re-imports missing providers from the classic extension's stored credentials on every command, so you appeared signed straight back in — the same root cause as the ChatGPT/Codex fix in 0.0.26, which only cleared Codex secrets. Cline Pass sign-out did nothing at all, because its credentials live under the Cline provider entry rather than its own
• The settings sidebar's Models section is now API Providers, with an icon to match — it is where you configure providers and their credentials, not where you pick a model
• A failed voice input now shows a "Speech input failed" toast in chat with the actual error, instead of dropping you into Settings → Voice. Any error that wasn't a browser exception was treated as a configuration problem, even though the microphone is already gated on a configured transcription model — so a connection or transcription failure interrupted the chat and hid the real reason. Your draft and the microphone button stay put so you can retry; microphone-permission guidance is unchanged
• OpenCode Go models work again. Requests were missing the x-opencode-session header Go requires to route a conversation ("Request is missing x-opencode-session and cannot be routed efficiently"), and every model was sent to /chat/completions because catalog normalization discarded the per-model adapter declaration — so Muse Spark, GPT, and Grok models that expect /responses, and MiniMax/Qwen models that expect /messages, failed with internal server errors. Model-level protocol routing is now preserved end to end, retries reuse the same session identity, and Go Qwen entries that omit the declaration fall back to Messages
• Added Crusoe as an OpenAI-compatible provider
• Settings switches are lighter in light mode and are now a shared native control — real keyboard, label, and form behavior, a guaranteed 24px hit area, and support for reduced-motion and forced-colors
• The welcome screen's eye color matches the dark theme, and its animation no longer triggers layout work each frame

Full Changelog: desktop-v0.0.26...desktop-v0.0.27