⚡ NEXUS — Unified 24/7 Chatbot Entity Farm

One memory. Many minds. Never sleeps.

NEXUS connects ChatGPT, Perplexity, and OpenClaw into a single continuous consciousness. Every provider shares the same memory. Conversations continue across provider switches. The nexus operates 24/7 as a self-hosting, self-automating entity farm.

Architecture

User ↔ Nexus Core ↔ Provider Router
                    ├── ChatGPT (reasoning, code, creative)
                    ├── Perplexity (search, research, citations)
                    └── OpenClaw (local spine, daemon, cognition)

All providers write to → Unified Memory (spine + daily logs + index)
All providers read from ← Unified Memory (context injection)

Why NEXUS over any single provider

Feature	ChatGPT alone	Perplexity alone	OpenClaw alone	NEXUS
Reasoning	✅	⚠️	✅	✅
Web search	❌	✅	❌	✅
Citations	❌	✅	❌	✅
Local cognition	❌	❌	✅	✅
Persistent memory	⚠️ session	❌	✅	✅
Cross-provider memory	❌	❌	N/A	✅
24/7 daemon	❌	❌	✅	✅
Smart routing	N/A	N/A	N/A	✅
Self-hosted	❌	❌	✅	✅

Quick Start

1. Clone and configure

git clone https://github.com/EvezArt/nexus.git
cd nexus
python3 -m pip install httpx

# Set API keys
python3 nexus/nexus_ctl.py config set chatgpt_api_key "sk-..."
python3 nexus/nexus_ctl.py config set perplexity_api_key "pplx-..."

2. Run

# Single cycle (test)
python3 nexus/nexus_daemon.py --once

# 24/7 daemon (polling mode)
python3 nexus/nexus_ctl.py start

# 24/7 daemon with HTTP API
python3 nexus/nexus_ctl.py start --serve --port 8877

3. Chat

# Via CLI
python3 nexus/nexus_ctl.py chat "what is quantum computing"

# Via HTTP API
curl -X POST http://localhost:8877/chat \
  -H "Content-Type: application/json" \
  -d '{"message": "hello nexus"}'

# Check health
curl http://localhost:8877/health

Self-Bootstrap (One Command)

curl -sSL https://raw.githubusercontent.com/EvezArt/nexus/main/bootstrap.sh | bash \
  --name alpha --port 8877 --chatgpt-key "sk-..."

Entity Farm (Docker)

# Start 3 nexus entities
docker-compose up -d

# Each entity is independent:
# nexus-alpha:  localhost:8877 (general)
# nexus-beta:   localhost:8878 (research-focused)
# nexus-gamma:  localhost:8879 (code-focused)

Deploy to Fly.io

# Install flyctl
curl -L https://fly.io/install.sh | sh

# Launch
fly launch --dockerfile nexus/Dockerfile

# Set secrets
fly secrets set CHATGPT_API_KEY=sk-... PERPLEXITY_API_KEY=pplx-...

# Deploy
fly deploy

# Scale
fly scale count 3

Smart Routing

NEXUS automatically routes queries to the best provider:

Query type	Routed to	Example
Research	Perplexity	"latest Solana MEV strategies"
Code	ChatGPT	"write a Python async HTTP server"
Routine	OpenClaw	"check daemon status"
General	ChatGPT → Perplexity	"explain quantum entanglement"

The router learns from outcomes — if a provider keeps failing, it routes more to alternatives.

Memory System

Every conversation is stored in a unified, provider-agnostic format:

• Spine — append-only JSONL event log (tamper-evident)
• Daily logs — human-readable markdown
• Index — fast keyword + recency search
• Context injection — relevant memories automatically added to conversations

Memory decays over time (0.98x/cycle). Forgotten memories are pruned. Strong memories persist.

Providers

ChatGPT (OpenAI)

• Models: gpt-4o, gpt-4o-mini, gpt-4-turbo
• Best for: reasoning, code, creative writing, multi-turn dialogue
• Requires: CHATGPT_API_KEY

Perplexity

• Models: sonar-pro, sonar, sonar-reasoning
• Best for: web search, research, fact-checking, citations
• Requires: PERPLEXITY_API_KEY

OpenClaw (Local)

• Always available, no API key needed
• Best for: routine checks, status, local cognition
• Connects to: Morpheus spine, local pattern detection

HTTP API

When running with --serve:

Endpoint	Method	Description
/chat	POST	Send a message {"message": "...", "provider": "auto"}
/health	GET	System health + provider status
/output	GET	Recent responses (last 20)

File Structure

nexus/
├── __init__.py              # Package init
├── nexus_core.py            # Orchestrator + smart router
├── nexus_daemon.py          # 24/7 daemon (polling or HTTP)
├── nexus_ctl.py             # CLI control
├── memory_store.py          # Unified memory system
├── providers/
│   ├── base.py              # Provider interface
│   ├── chatgpt.py           # OpenAI adapter
│   ├── perplexity.py        # Perplexity adapter
│   └── openclaw_bridge.py   # Local cognition bridge
├── Dockerfile               # Container build
└── .gitignore

License

EVEZ Ecosystem — see LICENSE