preload.js awaited two ipcRenderer.invoke calls before exposing
window.openswarm. Renderer code running in the gap saw the global
undefined: BrowserCard fell back to iframe mode (grey panel on sites
with X-Frame-Options), and chat-spawn calls to getAuthToken threw at
mount, crashing the renderer on Electron 40 + Windows.
Only bit Windows post-lazy-boot (ddd82ef): startBackend now loads main
while preload runs. Mac lacks the downstream native crash trigger;
pre-lazy Windows had idle IPC at preload time.
Replace awaits with sendSync against new ipcMain.on mirrors.
backendPort is assigned before any window exists, so sendSync at
preload time always finds a real value.
The render-process-gone handler called mainWindow.reload() to recover from
renderer crashes (e.g. WebGL / native / OOM in chat-spawn flows). On the
Electron 40 CastLabs build the reload re-registers a session/webview observer
that Chromium base/observer_list.h:318 refuses to add twice. The DCHECK
aborts the entire main process with exit 3, taking the app down whenever the
renderer dies.
Recreate the BrowserWindow instead: a fresh webContents has no duplicate
observers to add. The closed handler is identity-checked so the old window
teardown does not null out the new mainWindow reference.
When the 3-in-60s crash cap trips we now surface a native dialog (Reload /
Quit) instead of leaving a stuck dead window. Native dialog runs outside the
render-process-gone call stack, so its Reload button cannot itself reproduce
the DCHECK.
Bumps to 1.1.43 (re-cutting the deleted release draft).
+ webview popup dark-screen via new-window → exitFullscreen so parent stays interactive, ErrorSlime extracted to shared component and
rendered on all chat error cards (Network / Servers maxed / Plan limit / Subscription), Network issue classifier tightened to real errno
codes + retry CTA removed since edit_message-based replay was truncating successful tool history — copy now directs user to send-new-message
once WS auto-reconnects
- Add BUILTIN_MODELS for OpenAI (GPT-5.4/Mini/5.3-Codex), Google (Gemini 3 Pro/Flash, 2.5 Pro/Flash), and GitHub Copilot
- Add resolve_model_id_for_sdk() and resolve_aux_model() to route prefixed model IDs (cx/, gc/, gh/) through 9Router translator
- Add GET /agents/models endpoint returning available models based on live 9Router connection state
- Enable ENABLE_TOOL_SEARCH=auto for all providers so non-Claude models get full tool access (23 built-in + all MCPs)
- Add thinking block streaming support (ThinkingBlock in stream handler + AssistantMessage) for reasoning models
- Fix Gemini 3 thought-signature errors: use skip_thought_signature_validator per Google docs
- Fix WebSearch blocked_domains/allowed_domains empty-list rejection on Anthropic API
- Fix browser_agent text_parts UnboundLocalError on non-Claude models
- Fix auxiliary LLM calls (title gen, group meta, dashboard naming, view builder, browser agent) via resolve_aux_model
- Add Codex OAuth callback listener on port 1455 for ChatGPT Plus subscription connect
- Route Gemini OAuth through system browser since Google blocks embedded webviews
- Override Electron popup user-agent for OAuth
- Fix duplicate OAuth callback with idempotent completed_oauth tracking
- Fix Settings modal tab routing and stale warning banner
- Add API key inputs for OpenAI, Google, and OpenRouter
- Add MCP warning banner when selecting non-Claude model with many tools
- Force session fork on cross-provider model switch to prevent transcript corruption
- Disable GitHub Copilot subscription card (9Router poll issue, marked preview)
- Delete dead CopilotProvider import and unused CopilotAuthButton component
- nine_router.py: use Electron binary as Node.js fallback (ELECTRON_RUN_AS_NODE=1)
- electron/main.js: pass OPENSWARM_ELECTRON_PATH to backend
- OnboardingModal: show once for users without subscription (persisted via localStorage)
- settingsSlice: sync frontend default system prompt with backend
Eric