Loading...
AI Platform
UniversalAI.
Multi-model AI orchestration platform supporting parallel model execution, custom tools, and document intelligence.
3x LLMsPARALLEL STREAMS
InstantMULTILINGUAL TRANS
MonacoDEVELOPER SANDBOX
PrismaRELATIONAL POSTGRES
01 / Overview
Overview
UniversalAI is an advanced, production-grade AI orchestration and translation platform designed to unify state-of-the-art Large Language Models and specialized tools into a single, high-fidelity workspace. Built on a Next.js and TypeScript framework, it solves the challenge of single-model bias and tool fragmentation by running parallel execution streams from GPT-4o, Gemini Pro, and Claude 3.5 Sonnet side-by-side. The platform integrates a semantic Retrieval-Augmented Generation (RAG) system for document intelligence, alongside a dedicated AI agent workspace that generates, executes, and previews project files on the fly. By securing conversations through a robust PostgreSQL persistence layer and supporting instant multilingual communication, UniversalAI delivers a scalable, enterprise-ready environment for technical collaboration.
02 / Architecture Specs
Technical Specifications
| Component | Technology / Tool | Description |
|---|---|---|
| Core Architecture | Next.js 15 (React 19), TypeScript, Node.js v18+ | Provides high-throughput concurrent rendering and server-side model routing. |
| Database & storage | PostgreSQL (Neon / Supabase Serverless) | Relational database schema storing persistent user chats, settings, and histories. |
| ORM Layer | Prisma ORM | Provides type-safe querying, schema migrations, and structured relationship mapping. |
| Auth & Sessions | NextAuth.js (Auth.js) with JWT Session Cookies | Authenticates client sessions using JWT cookies and OAuth providers. |
| OpenRouter APIs | OpenRouter Aggregation Engine | Provides unified, low-latency API access to Anthropic Claude, Meta Llama, and OpenAI GPT models. |
| Gemini Integration | Google Gemini Flash & Pro Native APIs | Enables fast multi-modal image reasoning and massive context prompt checks. |
| Tool Orchestration | Model Context Protocol (MCP) | Secure client/server tool orchestration protocol to bind external services to LLM streams. |
| Animation & Styling | Radix UI Primitives, Tailwind CSS, Framer Motion | Powers premium responsive styling and fluid glassmorphism comparison transitions. |
| Payments & Gateways | Stripe & Razorpay API integrations | Handles webhook billing operations, invoices, and feature-gate variables. |
| Mobile Client | Kotlin & Android Studio (universe-ai-mobile) | Companion native Android client featuring direct API calls and local DB caching. |
03 / Platform Modules
Core Features
01
Concurrent Model Streams (Side-by-Side Comparison)
- Parallel Dispatch: Streams answers from up to 3 models (GPT-4o, Gemini, Claude 3.5 Sonnet) simultaneously side-by-side inside a unified layout.
- Resource Diagnostics: Monitors and displays real-time latency measurements, token consumption speeds, and API costs to analyze model efficiency.
02
Autonomous Developer Sandbox & Code Agent
- Folder Tree Generation: The AI coding assistant compiles complete repository configurations, writing structural files and code blocks on the fly.
- Interactive Preview: Renders live browser visual previews of generated components directly in a dedicated workspace frame.
03
Document Intelligence & Vector RAG Pipeline
- Fast Document Parsing: Parses user uploads (such as PDFs) locally and embeds key text passages into RAG contextual memory buffers.
- Grounded Reasoning: Injects document snippets back into prompt chains to assure AI responses are mathematically grounded in project data.
04
Real-Time Translation Hub & Feature Gates
- Preserved Syntax Translation: Dynamically translates active chat threads across languages without breaking code snippet blocks or Markdown elements.
- Usage Metering & billing: Stripe and Razorpay integrations monitor account limits and lock premium capabilities behind secure sub checks.
04 / Systems Flow
Architecture & System Flow
System Interaction Diagram
Table 01
users
Stores credentials, profile indicators, email handles, and dynamic Stripe subscription levels.
Table 02
sessions
Logs active user logins, verification keys, and NextAuth adapter parameters.
Table 03
chats
Collects user conversations, workspace selections, and custom instructions reference indices.
Table 04
messages
Holds prompt logs, model identifiers, outputs content, and response speed metrics.
Parallel Stream Workflow
- 1Prompt DispatchThe client sends a prompt along with model parameters to the Next.js API endpoint.
- 2Middleware FilteringAuth middleware decrypts session JWT cookies and validates user rate limits on Upstash Redis cache.
- 3Concurrent ExecutionThe endpoint maps the payload and dispatches parallel HTTP streams to OpenRouter (for Claude / Llama) and Gemini native API concurrently.
- 4Edge Token PipesAPI routes catch chunk-by-chunk stream chunks and pipe them directly back to the client using server-sent event (SSE) buffers.
- 5Real-time Comparative RenderClient hooks parse concurrent text stream pieces and feed them side-by-side into separate Framer Motion markdown boxes without blocking UI rendering.