Summary:
When using Kagi Assistant in Research Mode inside Orion, AI responses are
not fully rendered on screen. The response appears incomplete (cut off mid-
text) until I manually refresh the page (⌘R), after which the full response
becomes visible.
Steps to reproduce:
- Open Kagi Assistant (kagi.com/assistant) in Orion
- Select Research Mode
- Submit a query that produces a long, multi-step response
- Observe: response appears incomplete / cut off on screen
- Press ⌘R to refresh
- Observe: full response now visible
Actual behavior:
The response renders partially and stops. A manual refresh (⌘R) is required
to see the complete response. This makes Research Mode effectively unusable
without the workaround.
Environment:
- Orion version: [vul hier je versienummer in — te vinden via Orion → About Orion]
- macOS version: [vul hier je macOS-versie in]
- Hardware: [bijv. iMac M1 / MacBook Pro M3 Pro]
- Extensions active: [ja/nee — welke extensies heb je actief?]
- Kagi plan: Ultimate
Additional context:
This issue appears to be related to how Orion handles streaming responses
(Server-Sent Events / SSE). The same query in Firefox and Safari renders
completely without requiring a refresh. The issue occurs consistently in
Research Mode; standard (non-Research) Assistant responses may be less
affected.
This is particularly notable because Orion is Kagi's own browser — it would
be expected to have first-class support for Kagi's own streaming interfaces.
The full streamed response renders progressively and completely on screen,
without requiring a manual page refresh — as it does in Firefox and Safari.
macOS Tahoe 26.6 Orion Version 1.1.1 (149), WebKit 625.1.8
Tahoe (26)
