This project is no longer maintained. Some features may be non-functional; see Funnel instead.

Consensus

Multi-LLM evaluation platform querying 6+ models simultaneously with synthesized consensus responses and support for 11+ APIs through a unified routing layer.

ReactExpress.jsNode.jsGemini APIVite

Consensus interface showing multi-model query comparison

Consensus results with synthesized response generation

Overview

Consensus was a sandbox web app built to test multiple LLMs side-by-side in real time. It ran prompts across 6+ Gemini models at the same time, displayed streaming responses in parallel, and combined their answers to spot where models agreed or disagreed.

Highlights

Fast Streaming:

Streamed responses at 236 tokens/s with 2.5s p95 latency using Gemini 2.0 Flash.

Parallel Prompting:

Ran 6+ models concurrently with side-by-side LaTeX and code rendering in a collapsible layout.

Aggregated Outputs:

Used a React and Express backend to parse and merge outputs across multiple models.

Takeaways

Building Consensus made it clear that switching between separate dashboard tabs slows down day-to-day work. Those learnings directly shaped Funnel, shifting the focus toward an in-browser workflow that runs without leaving what you are reading.