Fuse Hackathon Autonomous AI Learning System

The system is anchored by a FastAPI entry point (server‑ml/app.py) that wires together controllers like PromptController, FeedbackController, and PersistenceController with their corresponding services. Each controller exposes a rich set of REST endpoints (e.g., /prompt/handle, /learning/start, /feedback/submit, /persistence/backup) that are documented in the repository’s README files. Models such as LLM, Prompt, CurriculumTree, and HallucinationMetrics encapsulate data structures and business rules, while services implement the core logic for learning loops, meta‑learning adaptation, hallucination detection, and reward calculation.

Data ingestion begins with a user prompt handled by PromptController, which validates input, logs the request, and forwards it to LearningLoopService. The service interacts with TaskGenerator and CurriculumGenerator to produce a curriculum, then schedules tasks via Scheduler. Throughout execution, AnalyticsService and VisualizationService collect metrics that are exposed via /analytics/* and /visualize/* endpoints. Persistence is managed by PersistenceService using versioned state objects, supporting backup, restore, and distributed sync operations.

Security guardrails are enforced through a middleware chain (request logging, security logging, rate limiting, JWT authentication, and Joi‑style validation) as described in API_DOCUMENTATION.md. This ensures that only sanitized, authenticated requests reach the ML backend, and that operational metrics are captured for monitoring. The architecture is designed for scalability: services can be horizontally scaled, and the distributed persistence layer allows state replication across multiple nodes.

The build topology uses a frontend (React/TypeScript) that proxies API calls through an Express server to the Python FastAPI backend. Development scripts (pip install -r requirements.txt, python app.py) and containerization hints are implied, enabling rapid iteration and deployment in a cloud or on‑prem environment.