Issue Description
When using Performance mode, responses are generated at a noticeably slow pace — characters/tokens appear one by one with significant delays between them, making the experience nearly unusable.
This issue occurs consistently in both the IDEA plugin and Qoder IDE.
Affected clients: IDEA plugin, Qoder IDE
Mode: Performance mode
Timezone: UTC+8 (Beijing Time)
Steps to Reproduce
- Switch to Performance mode
- Send any prompt during the affected time window
- Observe that the response trickles out character by character instead of streaming normally
Expected Behavior
Response streams at a normal, continuous pace regardless of time of day.
Actual Behavior
Response output is severely throttled — tokens appear one at a time with visible delays, as if the backend is under heavy load or rate-limited.
Screenshots / Screen Recordings
Operating System
Windows11 64G i7 13600KF NVIDA Geforce RTX 4070 SUPER
Current Qoder Version (Menu → About Qoder → Copy)
IdealJ 64 - 2026.1
Version: 0.12.1
VSCode Version: 1.106.3 (user setup)
Commit: 8528fc5abcf430919311b276cc3f33a60a270e06
Date: 2026-04-08T11:07:21.893Z
Electron: 37.7.0
Chromium: 138.0.7204.251
Node.js: 22.20.0
V8: 13.8.258.32-electron.0
OS: Windows_NT x64 10.0.26200
