v3.1.0 — Qwen Coding Plan & Stream Stall Indicator

What's New in v3.1.0

Stream Stall Indicator

• Animated spinner (⣾ waiting for response... 5s) when streaming pauses for 2+ seconds
• Works across all 3 providers (Claude, OpenAI-compat, Ollama)
• Clears automatically when the next chunk arrives

Qwen Coding Plan Support

• Auto-detects sk-sp- Coding Plan keys during /key qwen
• Region picker: International (Singapore) or China (Beijing)
• Dedicated endpoint: coding-intl.dashscope.aliyuncs.com/v1
• Auto-remaps models for Coding Plan compatibility (e.g. qwen3-max → qwen3-max-2026-01-23)
• Third-party models included: GLM-5, Kimi K2.5
• Lite ($3/mo) and Pro ($15/mo) tiers supported

Qwen API Fixes

• Fixed all model aliases to use correct DashScope IDs (qwen3-prefixed)
• Removed invalid model IDs (qwen-long-latest, qwen3.5-27b)
• Provider-aware request body — parallel_tool_calls only sent to OpenAI

UX Improvements

• /key now shows "paste new key to replace, or type 'remove'" when updating existing keys
• API error messages now show the actual error from the provider
• Debug line on 400 errors shows model ID + endpoint URL

Stats

• 232 tests passing
• 10 source files, ~7,800 lines
• 63 slash commands
• 26 tools
• 7 API providers