Skip to content

usage.md

Latest commit

 

History

History
101 lines (74 loc) · 4.02 KB

usage.md

File metadata and controls

101 lines (74 loc) · 4.02 KB
Raw

GhostAI: Anonymous Multi-Model Intelligence & ZK Proof of Existence

GhostAI Landing Screen

GhostAI is a high-performance, parallel-streaming chat environment that leverages the Midnight blockchain for Zero-Knowledge Proofs of Existence. This document outlines the technical requirements, local configuration, and deployment procedures.

1. Prerequisites

Before initializing the GhostAI stack, ensure the following dependencies are installed and configured:

  • Node.js: v18.17.0 or higher (v20+ recommended for Next.js Turbopack).
  • Docker Desktop: Required to run the local Midnight node, ZK proof server, and Compact compiler container.
  • Midnight Lace Wallet: Install the browser extension to interact with the DApp locally. Ensure it is configured to point to the local devnet.
  • Git: For repository cloning and version control.

2. Local Environment Configuration

Repository Setup

Clone the repository and install the Next.js dependencies:

git clone https://github.com/your-org/ghost-ai.git
cd ghost-ai/ghost-ai-ui
npm install

Environmental Variables

GhostAI utilizes a dynamic multiplexing system for API requests to bypass rate limits during multi-model parallel comparisons. Create a .env.local file in the ghost-ai-ui directory:

touch .env.local

Populate it with the provided OpenRouter API keys. Ensure these keys are never committed to source control.

# OpenRouter Proxy Keys - DO NOT COMMIT
OPENROUTER_API_KEY_1=sk-or-v1-xxxxxx
OPENROUTER_API_KEY_2=sk-or-v1-xxxxxx
OPENROUTER_API_KEY_3=sk-or-v1-xxxxxx

# Application State
NEXT_PUBLIC_ENVIRONMENT=development

Warning

Key Safety: The .env.local file must remain in your .gitignore. If these keys are exposed, the parallel streaming engine will face immediate 429 Rate Limit floods.

3. Spinning Up the Stack

GhostAI requires both the blockchain infrastructure and the frontend client to be running simultaneously.

Step 3A: Initialize Midnight Devnet

Navigate to the Midnight local environment directory to spin up the local blockchain ledger and proof server.

cd midnight-local-dev
docker compose up -d

Verify that the containers (midnight-node, proof-server) are healthy via Docker Desktop.

Step 3B: Compile the Smart Contract

Navigate to the smart contract store and compile the Compact circuit.

cd ../hash-store
npm install
npm run compile

This generates the ZK circuit artifacts required by midnightBridge.ts to seal conversation hashes.

Step 3C: Launch the GhostAI Client

Return to the Next.js frontend and start the development server using Turbopack for optimized HMR.

cd ../ghost-ai-ui
npm run dev

Navigate to http://localhost:3000 to access the GhostAI Landing Environment.

GhostAI Live Dashboard

4. Cloud Deployment (Production)

To deploy GhostAI for live hackathon judging, the architecture splits between a static/edge frontend and a robust backend proxy.

Vercel Frontend Deployment

  1. Connect your GitHub repository to Vercel.
  2. Set the Root Directory to ghost-ai-ui.
  3. In the Environment Variables panel, map your OpenRouter API keys exactly as they appear in .env.local:
    • OPENROUTER_API_KEY_1
    • OPENROUTER_API_KEY_2
    • OPENROUTER_API_KEY_3
  4. Deploy the application. Vercel will automatically route the Next.js API endpoints (/api/chat/route.ts) to Edge Functions, ensuring minimal latency for the streaming responses.

Render / Railway Network Proxy Configuration (Optional / Advanced)

If Vercel's 10-second serverless timeout becomes a bottleneck for heavy 120B parameter models during Comparison Mode, deploy a standalone proxy:

  1. Extract /api/chat to a standalone Node.js/Express server.
  2. Deploy the Express server to Render or Railway.
  3. Configure the Render/Railway environmental variables with the OpenRouter key multiplexing matrix.
  4. Update GhostAiDashboard in page.tsx to point the stream fetches to your new Render URL instead of the local Next.js route.