Copyright © 2026 Akimitsujiro. All rights reserved.
This repository is published strictly for technical reference and portfolio demonstration purposes.
- No Unauthorized Distribution: Unauthorized cloning, downloading, or redistribution of this source code for public or commercial use is strictly prohibited.
- No Direct Execution: This is not a "plug-and-play" template for public use.
- Permission Required: Any usage, adaptation, or deployment of the logic found within this repository requires explicit written consent from the author.
EtherealAI is a high-performance, full-stack platform designed to orchestrate complex interactions with dynamic AI personas and autonomous agents. By utilizing an advanced multi-model routing architecture, EtherealAI delivers ultra-responsive, context-aware AI companions with persistent memory.
Currently, the core inference engine is undergoing a strategic migration to NVIDIA NIM to harness the robust capabilities of state-of-the-art models like GLM-5.1 and DeepSeek V4 for high-concurrency, agentic reasoning tasks.
To support real-time, multi-turn conversations and background automated tasks (e.g., semantic memory compression, sentiment analysis), EtherealAI requires a robust, low-latency infrastructure.
- GLM-5.1 Integration: Deployed for complex agentic decision-making, self-correction, and strict adherence to character system prompts.
- DeepSeek V4 Integration: Leveraged for processing massive context windows, ensuring deep narrative consistency during extended user-persona chat sessions.
- High-Concurrency Architecture: Our pipeline relies on parallel API calls for safety-checking, context retrieval, and response generation, which strictly necessitates a 200 RPM (Requests Per Minute) allowance to prevent bottlenecking during development and testing.
- Dynamic AI Personas: Create, customize, and interact with unique AI characters with complex system prompts.
- Real-Time Streaming Inference: Low-latency token streaming optimized for seamless user experience.
- Cross-Platform Ecosystem: Built-in Discord bot interactions synced directly with web platform accounts.
- Scalable Credit System: Integrated user economy with Stripe & PayOS webhooks for premium tiers.
- Comprehensive Admin Hub: Full oversight over user roles, API status, and community reports.
- Frontend/Backend: Next.js 14 (App Router) & TypeScript.
- Database: Prisma ORM for relational data management (Characters, Chats, Users, Transactions).
- Authentication: NextAuth.js with secure turnstile verification.
- Styling: Tailwind CSS with custom Glassmorphism UI components.
- AI Inference Orchestration: NVIDIA NIM API (Primary) combined with hybrid local tools (OpenClaw/Gemma4).
Developed by Akimitsujiro. This project is a private intellectual property and is displayed here for the NVIDIA NIM developer review process.