Orythm — AI Observability Platform

Orythm is a full-stack AI observability and analytics platform designed to monitor, analyze, and visualize Large Language Model (LLM) interactions in real time.

The platform captures:

• Prompt history
• Token usage
• Latency metrics
• Model performance
• Tool-call analytics
• Interaction timelines
• Iteration tracking

Built using:

• Spring Boot
• React
• PostgreSQL
• Docker
• Gemini API integration

---

Features

AI Interaction Analytics

Track and analyze:

• Input prompts
• Generated responses
• Tokens in / tokens out
• Response duration
• Start and end timestamps
• Model metadata

Interactive Dashboard

Visualize:

• Latency trends
• Token consumption
• Performance metrics
• Prompt iteration history
• Analytics summaries

REST API

Expose APIs for:

• Uploading interaction data
• Querying analytics
• Tracking model performance
• Integrating external AI systems

Dockerized Deployment

Run the entire platform locally using Docker Compose.

---

Tech Stack

Layer	Technology
Backend	Spring Boot (Java 21)
Frontend	React + TypeScript
Database	PostgreSQL
Infrastructure	Docker + Docker Compose
AI Integration	Gemini API

---

Architecture

User Prompt
    ↓
Gemini API
    ↓
Spring Boot Backend
    ↓
PostgreSQL Analytics Storage
    ↓
React Dashboard

---

Running Locally

Prerequisites

• Docker
• Docker Compose

Start Application

docker compose up --build

Access Application

http://localhost:8080

---

Future Enhancements

• Multi-model comparison
• Real-time streaming analytics
• Cost estimation dashboard
• Prompt versioning
• AI evaluation metrics
• Agent telemetry support

---

Project Goal

Orythm was built to explore AI observability, prompt analytics, and LLM performance monitoring using modern full-stack engineering practices.

The platform demonstrates:

• AI system telemetry
• LLM analytics pipelines
• Full-stack architecture
• Dockerized deployment
• API-driven observability workflows