Argument Evaluator

A tool for structured argument analysis and critical thinking evaluation.

Overview

The Argument Evaluator analyses written arguments through a multi-stage pipeline:

1. Argument Extraction - Extracts structured components (premises, conclusion, context)
2. Critical Thinking Checks - Runs clarity, consistency, logical support, epistemic support, and counterargument checks
3. Scoring - Produces a 0-100 scorecard across six rubrics
4. Human-Readable Summary - Provides strengths, weaknesses, hidden assumptions, and rewrite suggestions

Key Features

• Validity vs Factual Support Separation - Distinguishes between logical validity and empirical evidence
• Evidence Hooks - Flags empirical and causal claims that require verification
• Schema-First Design - Stable JSON API for easy integration
• v0-Friendly Stack - Built with Next.js App Router, TypeScript, Tailwind, and shadcn/ui

Project Structure

.
├── apps/
│   └── web/              # Next.js frontend
├── services/
│   └── engine/           # Python FastAPI backend
├── schemas/              # JSON schemas
├── samples/              # Example inputs and outputs
├── specs/                # Product specifications
├── AGENTS.md             # How to run services
├── IMPLEMENTATION_PLAN.md # Task queue
└── loop.sh               # Ralph Wiggum loop workflow

Quick Start

1. Set Up the Engine

cd services/engine
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate
pip install -e ".[dev]"

2. Set Up the Web UI

pnpm install

3. Run Both Services

Terminal 1 (Engine):

cd services/engine
source venv/bin/activate
uvicorn app.main:app --reload --port 8000

Terminal 2 (Web):

cd apps/web
pnpm dev

Then open http://localhost:3000

Running Tests

cd services/engine
source venv/bin/activate
pytest

Development Workflow

This project uses the Ralph Wiggum loop:

./loop.sh plan   # Planning phase
./loop.sh build  # Building phase

See IMPLEMENTATION_PLAN.md for the current task queue.

Documentation

• AGENTS.md - How to run services locally
• specs/argument-evaluator.md - Product specification
• .claude/CLAUDE.md - Project architecture and conventions
• schemas/ - JSON schemas for API contracts

Licence

MIT