Skip to content

TrendySloth1001/argumentbot

BranchesTags

Repository files navigation

# MrArgue

AI Debate Simulator - A real-time adversarial debate system where two Large Language Models argue opposing positions on any topic, scored by an impartial AI judge.

License: MIT NestJS Flutter Ollama Release

Screenshots

Home Debate History
Home Debate History
Feed Settings
Feed Settings

Overview

MrArgue pits two locally-hosted LLMs against each other in structured debates. Each model is constrained by strict rules that force adversarial reasoning: explicit claims, direct rebuttals, and pointed counter-questions. An AI judge scores each turn based on claim quality, directness, and attack strength.

Key Differentiators

  • Adversarial by design: Prompts enforce falsifiable claims and penalize vague abstractions
  • RAG-enhanced: Retrieval-augmented generation provides factual context
  • Dual scoring modes: AI judge or algorithmic heuristics
  • Entertainment value: AI models roast each other with witty comebacks

System Architecture

System Architecture

Features

Implemented

Category Feature Description
Debates Multi-Model Debates Llama 3.2 (PRO) vs Gemma 2:2b (CON)
Streaming Responses Server-Sent Events for real-time generation
RAG Context Embedding-based fact retrieval
Debate History Persistent storage with replay
Scoring AI Judge LLM-based analysis with detailed metrics
Algorithmic Judge Heuristic scoring for instant results
Power Bar Visual cumulative score display
Auth JWT Authentication Stateless token-based auth
User Registration Email/password with Argon2 hashing
Profile Management Avatar selection, nickname/age
Persistent Auth "Remember Me" session persistence (7 days)
Social Community Feed Posts with like functionality
Debate Sharing Share results link generation
UI Dark Theme Black + neon green aesthetic
How It Works Interactive explainer with LaTeX formulas

Roadmap

Priority Feature Status
High User vs AI Mode Planned
High Topic Browser Planned
Medium Social Sharing Planned
Medium Debate Rematch Planned
Medium Leaderboard Planned
Low Voice Input Planned
Low Multi-language Planned

Technology Stack

Backend

Component Technology Version Purpose
Runtime Node.js 20+ JavaScript runtime
Framework NestJS 11.x API framework
ORM Prisma 5.x Database access
Database SQLite 3.x Data persistence
Auth Passport + JWT - Authentication
Caching cache-manager 7.x Response caching
NLP natural 8.x Algorithmic scoring

Frontend

Component Technology Version Purpose
Framework Flutter 3.10+ Cross-platform UI
HTTP http 1.6.x API communication
Storage flutter_secure_storage 10.x Token storage
Math flutter_math_fork 0.7.x LaTeX rendering
Navigation google_nav_bar 5.x Bottom navigation

AI Infrastructure

Component Technology Purpose
LLM Runtime Ollama Local model hosting
Proponent Model Llama 3.2 Fast, consistent reasoning
Opponent Model Gemma 2:2b Creative, divergent arguments
Embedding Model nomic-embed-text Vector embeddings for RAG

Scoring Algorithms

AI Judge Scoring

The AI judge evaluates each turn on three dimensions with vagueness penalties:

Persuasiveness = (C_claim + C_answer + C_attack) / 3 - V_penalty
Metric Range Criteria
C_claim 0-100 Falsifiable claim with evidence
C_answer 0-100 Direct response to opponent's question
C_attack 0-100 Specific flaw identification in opponent's argument
V_penalty -30 each Penalty for undefined abstractions

Algorithmic Scoring

Instant heuristic-based scoring:

Score = w1 * L_arg + w2 * K_rel + w3 * C_struct
Factor Weight Description
L_arg 0.3 Argument length (optimal: 50-150 words)
K_rel 0.4 Keyword relevance to topic
C_struct 0.3 Structural compliance (headers present)

Power Bar Calculation

Cumulative persuasiveness ratio:

R_pro = SUM(P_pro) / (SUM(P_pro) + SUM(P_con)) * 100%

RAG Similarity

Cosine similarity for context retrieval:

similarity(q, d) = (q . d) / (||q|| * ||d||)

Database Schema

Core Tables

Users

Column Type Constraints Description
id UUID PRIMARY KEY Unique identifier
email VARCHAR UNIQUE, NOT NULL Login email
password VARCHAR NOT NULL Argon2 hash
avatarUrl VARCHAR NULLABLE Profile image URL
nickname VARCHAR NULLABLE Display name
age INTEGER NULLABLE User age
createdAt TIMESTAMP DEFAULT NOW Registration time

Debates

Column Type Constraints Description
id UUID PRIMARY KEY Unique identifier
topic VARCHAR NOT NULL Debate topic
status ENUM NOT NULL ACTIVE, FINISHED
userId UUID FOREIGN KEY Owner reference
createdAt TIMESTAMP DEFAULT NOW Start time

DebateTurns

Column Type Constraints Description
id UUID PRIMARY KEY Unique identifier
debateId UUID FOREIGN KEY Parent debate
speaker ENUM NOT NULL MODEL_A, MODEL_B
content TEXT NOT NULL Turn content
modelName VARCHAR NOT NULL LLM model used
analysis JSON NULLABLE Scoring data
timestamp TIMESTAMP DEFAULT NOW Turn time

API Reference

Authentication Endpoints

Method Endpoint Body Response
POST /auth/register {email, password} {token, user}
POST /auth/login {email, password} {token, user}
GET /auth/verify - {user}
POST /auth/onboarding {nickname, age} {user}
PUT /auth/profile {avatarUrl} {user}

Debate Endpoints

Method Endpoint Body Response
GET /debate - [Debate]
GET /debate/:id - Debate
POST /debate/start {topic} Debate
POST /debate/:id/turn {scoringMode} Debate
GET /debate/:id/turn/stream - SSE Stream

Feed Endpoints

Method Endpoint Body Response
GET /feed - [Post]
POST /feed {content, debateId?} Post
POST /feed/:id/like - {liked}

Installation

Prerequisites

Requirement Version Purpose
Node.js 20+ Backend runtime
npm 10+ Package manager
Flutter 3.10+ Frontend framework
Ollama Latest LLM runtime

1. Clone Repository

git clone https://github.com/TrendySloth1001/argumentbot.git
cd argumentbot

2. Backend Setup

cd backend

# Install dependencies
npm install

# Configure environment
cp .env.example .env
# Edit .env with your settings

# Initialize database
npx prisma generate
npx prisma migrate dev

# Start development server
npm run dev

3. Frontend Setup

cd frontend

# Install dependencies
flutter pub get

# Configure API URL
# Edit lib/core/config/api_config.dart

# Run application
flutter run

4. Ollama Setup

# Install Ollama from https://ollama.ai

# Pull required models
ollama pull llama3.2
ollama pull gemma2:2b
ollama pull nomic-embed-text

# Verify installation
ollama list

Docker Deployment

Build Backend Image

cd backend
docker build -t argumentbot-backend:latest .

Run Container

docker run -d \
  --name argumentbot-api \
  -p 3000:3000 \
  -e DATABASE_URL="file:./dev.db" \
  -e JWT_SECRET="your-secret-key" \
  -e OLLAMA_API_URL="http://host.docker.internal:11434" \
  argumentbot-backend:latest

Docker Compose

cd backend
docker-compose up -d

Environment Variables

Backend Configuration

Variable Required Default Description
DATABASE_URL Yes - SQLite connection string
JWT_SECRET Yes - JWT signing secret
OLLAMA_API_URL No http://localhost:11434 Ollama API endpoint

Project Structure

Directory Structure

⚠️ Known Issues

  • Karaoke Sync: The text highlighting in Karaoke mode involves estimating word duration based on character count. This may occasionally drift slightly from the actual audio playback, especially with faster voice models.
  • Emulator Audio: Audio playback latency can be higher on Android Emulators/iOS Simulators compared to physical devices.
  • First-Time Load: The first synthesis request for a specific phrase may have slight latency as the backend initializes the stream; subsequent requests are cached.

Contributing

  1. Fork the repository
  2. Create a feature branch: git checkout -b feature/new-feature
  3. Commit changes: git commit -m 'Add new feature'
  4. Push to branch: git push origin feature/new-feature
  5. Submit a Pull Request

License

This project is licensed under the MIT License. See LICENSE for details.

About

it just crossed my mind and i build it may be some day it will be helpful lol

Topics

text-to-speech ai-agents llms

Resources

Readme
Activity

Stars

1 star

Watchers

0 watching

Forks

0 forks
Report repository

Releases

1 tags

Packages

 
 
 

Contributors

Languages