- Custom Fine-Tuned Cross Encoder: fine-tuned model to detect logical coherence and argument strength.
- Logical Analysis: Deconstructs arguments into Claim → Evidence.
- Bias Detection: Identifies cognitive traps like Sunk Cost Fallacy.
- RAG Context: Checks new decisions against your personal archive.
- Qdrant Vector DB: Semantic search for similar past dilemmas.
- Real-time Processing: Asynchronous analysis with polling.
Backend
- FastAPI: High-performance async API.
- PyTorch (Last Mile): Custom Cross-Encoder for logic scoring.
- LangChain: Orchestration for RAG and LLM flows.
Frontend
- React + Vite: Fast, modern UI.
- Tailwind CSS: Sleek, responsive design.
- Framer Motion: Smooth animations.
Infrastructure
- Docker Compose: Full stack orchestration.
- Qdrant: High-performance vector search.
- Nginx: Production-grade reverse proxy.
make deploy-prodAccess at: http://localhost
make build
make run- Use Cases - Examples and scenarios.
- Deployment - Setup guide.
- Architecture - System design.