Arena

🌐 Language: English | 中文

Arena (Multi-Engine Comparison)

Arena is a deep analytics dashboard that lets you run the same prompt across multiple engines side-by-side and compare outputs.

How to open

• Sidebar ⋯ menu → Arena

What it shows

Section	Description
Diff view	Side-by-side comparison of two engine outputs with word-level diff highlighting
Judge	AI-judged comparison of which output is better
Select two	Choose which two engines/sessions to compare
Left / Right panels	The two selected outputs shown in parallel

How it works

1. Select two engines (e.g., Direct vs Claude Code) or two sessions.
2. Send a prompt — the same prompt goes to both engines simultaneously.
3. Compare outputs — Arena shows word-level diff, token counts, latency, and cost.
4. AI judge — optionally ask the AI to judge which output is better.

Source files

File	Purpose
src/renderer/arena.ts	Arena dashboard rendering, diff, judge logic
src/renderer/styles.css	Arena-specific styles
src/shared/i18n.ts	arena.* keys

Related

• Engines — the three engines and how they differ
• Session-Branching — branch a session to another engine