fix(copilot): persist thinking blocks on page refresh

[waleedlatif1]

Summary

• Use navigator.sendBeacon in beforeunload to persist in-progress messages (including thinking blocks) during page refresh/close
• Flush batched streaming updates before beacon persistence so contentBlocks are up to date
• abortMessage uses beacon instead of fire-and-forget fetch when page is unloading
• Fix double-digit ordered list clipping in thinking block (pl-6 → pl-8)

Type of Change

• Bug fix

Testing

Tested manually

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
docs	Skipped		Feb 11, 2026 6:46am

[greptile-apps]

Greptile Summary

This PR fixes the loss of thinking blocks and in-progress copilot messages during page refresh/close. It uses navigator.sendBeacon — which the browser guarantees to dispatch even during page teardown — to persist the current conversation state. A secondary cosmetic fix corrects ordered-list padding in the thinking block UI.

• persist.ts: Extracts a shared buildPersistBody helper and adds persistMessagesBeacon, which wraps navigator.sendBeacon with a Blob payload. Using a Blob with type: 'application/json' is the correct way to set Content-Type since sendBeacon does not accept custom request headers
• store.ts: Registers a beforeunload listener at module load time that sets _isPageUnloading = true, flushes any RAF-batched streaming updates via flushStreamingUpdates, then immediately calls persistMessagesBeacon. The abortMessage action is extended to switch from async fetch to sendBeacon when unloading, and to preserve activeStream in session storage (so the stream can be resumed after the refresh)
• thinking-block.tsx: Changes [&_ol]:!pl-6 → [&_ol]:!pl-8 in both the streaming (SmoothThinkingText) and static post-stream content divs so that double-digit ordered-list markers are no longer clipped

Confidence Score: 5/5

Safe to merge — no blocking issues found; all changes are focused bug fixes

The sendBeacon approach is the correct browser-native pattern for page-unload persistence. The beforeunload handler is synchronous, wrapped in a no-throw try/catch, and only fires when there is an active streaming message. flushStreamingUpdates correctly drains the RAF queue before persisting, ensuring thinking blocks are captured. The abortMessage path correctly suppresses the continue option and preserves the active stream on unload. The pl-6 to pl-8 change is a simple, safe cosmetic fix. No P0 or P1 issues were identified.

No files require special attention

Important Files Changed

Filename	Overview
apps/sim/stores/panel/copilot/store.ts	Adds beforeunload handler that flushes batched streaming updates and persists in-progress messages via sendBeacon; extends abortMessage to use beacon and preserve activeStream during page unload
apps/sim/lib/copilot/messages/persist.ts	Adds persistMessagesBeacon using navigator.sendBeacon with a JSON Blob; extracts shared buildPersistBody helper used by both the fetch and beacon paths
apps/sim/app/workspace/[workspaceId]/w/[workflowId]/components/panel/components/copilot/components/copilot-message/components/thinking-block/thinking-block.tsx	Fixes double-digit ordered list clipping in thinking block by changing pl-6 to pl-8 in both streaming and static content views

Sequence Diagram

sequenceDiagram
    participant Browser
    participant CopilotStore
    participant Server as Server (update-messages)

    Note over Browser,CopilotStore: User refreshes or closes tab while streaming

    Browser->>CopilotStore: beforeunload fires
    CopilotStore->>CopilotStore: _isPageUnloading = true
    alt isSendingMessage && currentChat
        CopilotStore->>CopilotStore: flushStreamingUpdates(setState)
        Note over CopilotStore: Drains RAF-batched contentBlocks into store
        CopilotStore->>Server: navigator.sendBeacon(blob)
        Note over Server: Browser queues beacon — delivered after teardown
    end

    Note over Browser,CopilotStore: abortMessage() called (e.g. via cleanup on unmount)

    CopilotStore->>CopilotStore: isPageUnloading() === true
    CopilotStore->>CopilotStore: suppressContinueOption = true
    CopilotStore->>CopilotStore: flushStreamingUpdates(set) — no-op if already flushed
    CopilotStore->>CopilotStore: Preserve activeStream in sessionStorage
    Note over CopilotStore: Stream can be resumed after page reload
    CopilotStore->>Server: navigator.sendBeacon(blob)
    Note over Server: Persists thinking blocks + partial content

Loading

_{Reviews (2): Last reviewed commit: "fix(copilot): persist thinking blocks on..." | Re-trigger Greptile}

[greptile-apps]

_{9 files reviewed, 1 comment}

_{Edit Code Review Agent Settings | Greptile}

[greptile-apps]

Using any type violates TypeScript conventions - consider using proper types from Streamdown's component prop types

Suggested change

	h2: ({ children }: any) =>
	h2: ({ children }: { children?: React.ReactNode }) =>

Context Used: Context from dashboard - TypeScript conventions and type safety (source)

_{Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is OFF. To automatically fix reported issues with Cloud Agents, enable Autofix in the Cursor dashboard.}