██╗     ██╗     ███╗   ███╗██████╗ ██╗███╗   ██╗ ██████╗
██║     ██║     ████╗ ████║██╔══██╗██║████╗  ██║██╔════╝
██║     ██║     ██╔████╔██║██████╔╝██║██╔██╗ ██║██║  ███╗
██║     ██║     ██║╚██╔╝██║██╔══██╗██║██║╚██╗██║██║   ██║
███████╗███████╗██║ ╚═╝ ██║██║  ██║██║██║ ╚████║╚██████╔╝
╚══════╝╚══════╝╚═╝     ╚═╝╚═╝  ╚═╝╚═╝╚═╝  ╚═══╝ ╚═════╝
                 " Local LLM Battleground "

Compare local LLMs head-to-head in a fast, keyboard-first terminal arena.

llmring brings side-by-side model evaluation to your terminal with streaming responses, blind voting, and Elo-based rankings.

Why Use It

• Compare 2-6 local models on the same prompt
• Stream responses in real time to judge quality and speed
• Run blind evaluations to reduce model-name bias
• Vote winners instantly and build an Elo leaderboard
• Stay in flow with full keyboard navigation

Features

• Terminal-native UX: interactive, chat-style TUI
• Live generation: token streaming while models respond
• Blind mode: hide model names during evaluation
• Flexible voting: vote by label (A-F) or mark a tie
• Persistent stats: sessions, votes, and Elo saved locally
• Prompt loop: iterate quickly with next/regenerate controls

Prerequisites

• Node.js 14 or later
• Ollama installed and running

Install

git clone https://github.com/12britz/llmring.git
cd llmring
npm install
npm run build
npm link

Quick Start

# Launch the TUI
llmring start

# Optional helpers
llmring list
llmring leaderboard

Usage Flow

1. Run llmring start
2. Select models from the top bar (click to cycle)
3. Enter a prompt in the input box
4. Press Ctrl+E to generate responses
5. Vote (A/B) or mark tie (G/D)
6. Continue with N (next) or R (regenerate)

Controls

Key	Action
Ctrl+E	Send prompt
Ctrl+Tab	Cycle focus
A / B	Vote for model
G	Tie (good)
D	Tie (bad)
R	Regenerate responses
N	Next / Reset
Ctrl+C	Quit

Data Storage

Local data is stored in:

• ~/.llmring/sessions.json
• ~/.llmring/votes.json
• ~/.llmring/elo.json

Interface Layout

┌─────────────────────────────────────────────────────────────┐
│                        LLMRing TUI                         │
├──────────────┬─────────────────────────────────────────────┤
│ Models       │ Response Panels                             │
│ Sidebar      │ ┌─────────────────────────────────────────┐ │
│              │ │ [A] Model A                             │ │
│ [x] llama3   │ │ Response from model A...                │ │
│ [ ] qwen2.5  │ └─────────────────────────────────────────┘ │
│ [x] mistral  │ ┌─────────────────────────────────────────┐ │
│ [ ] gemma2   │ │ [B] Model B                             │ │
│              │ │ Response from model B...                │ │
│              │ └─────────────────────────────────────────┘ │
├──────────────┴─────────────────────────────────────────────┤
│ [Type your prompt here...]                                 │
├─────────────────────────────────────────────────────────────┤
│ [A-F] Vote | [T] Tie | [N] Next | [R] Regen | [B] | [Q]   │
└─────────────────────────────────────────────────────────────┘

License

MIT