Skip to content

bigdatasciencegroup/debate-spoon

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

16 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Adversarial Science Tribunal

Four AI agents tear apart research papers in real-time debate, then store the verdict on Neo blockchain.

The Problem

Peer review is broken. Papers take months to review, reviewers are overworked, and bad science still slips through. We built a system that catches methodological issues instantly - not to replace human reviewers, but to give researchers immediate feedback before they even submit.

How It Works

Upload a paper. Four specialized agents analyze it in parallel:

  • The Skeptic - hunts for alternative explanations and confounding variables
  • The Statistician - catches p-hacking, underpowered studies, and statistical misuse
  • The Methodologist - evaluates experimental design, controls, and reproducibility
  • The Ethicist - flags conflicts of interest, consent issues, and disclosure problems

They don't just analyze - they debate. Each agent challenges the others' findings. You can jump in anytime, ask questions, push back on their conclusions, or request clarification.

When ready, request a verdict. The system scores your paper 0-100, lists critical issues, and permanently records everything on Neo blockchain.

Tech Stack

SpoonOS SDK - Orchestrates the multi-agent system. Each agent runs independently with specialized prompts, then coordinated for cross-examination.

Neo N3 Blockchain - Every verdict gets recorded on-chain. Paper hash, score, critical issues - all immutable and timestamped.

Mem0 - Long-term memory layer with semantic embeddings. Find similar papers, track issues across your research, spot patterns.

ElevenLabs - Voice synthesis for the debate. Each agent has a distinct voice. Supports real-time voice interaction.

Language Support - Full English and Chinese. Agents analyze Chinese papers in Chinese with localized terminology.

FastAPI + Next.js - Python backend for orchestration, React frontend with real-time updates.

Setup

# Backend
cd backend
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
cp .env.example .env
uvicorn src.api.main:app --reload --port 8000

# Frontend
cd frontend
npm install
npm run dev

Environment Variables

ANTHROPIC_API_KEY=     # Claude for agent reasoning
ELEVENLABS_API_KEY=    # Voice synthesis
NEO_PRIVATE_KEY=       # Neo blockchain signing
MEM0_API_KEY=          # Long-term memory

API

POST /api/interactive/start-pdf   Start with PDF upload
POST /api/interactive/start       Start with text
POST /api/interactive/{id}/message  Send message during debate
POST /api/interactive/{id}/request-verdict  Get final verdict
GET  /api/verdicts                 List all verdicts
GET  /api/verdicts/{id}            Verdict details
GET  /api/verdicts/by-paper        Papers grouped by version
POST /api/voice/synthesize         Text to speech
POST /api/voice/transcribe         Speech to text

MIT License

Releases

No releases published

Packages

No packages published

Languages

  • TypeScript 50.9%
  • Python 44.6%
  • PowerShell 1.8%
  • CSS 1.1%
  • C 1.0%
  • Shell 0.6%