TradeMind — Onchain AI Trading Agent

A personal onchain AI agent that learns from your trading behavior, stores persistent memory, and evolves its responses over time.

Not a chatbot. A system with memory.

Quick Start

npm install
npm run dev

Open http://localhost:3000. No API key required — adaptive fallback works out of the box.

Optional: Add AI

Edit .env.local:

OPENAI_API_KEY=sk-...
OPENAI_BASE_URL=https://api.openai.com/v1
OPENAI_MODEL=gpt-4o-mini

Architecture

types/index.ts              All TypeScript interfaces
lib/storage/index.ts        Mock 0G Storage (file-backed KV per user)
lib/agent/behavior.ts       Pattern extraction + behavior learning
lib/agent/reasoning.ts      AI prompting layer (OpenAI + fallback)
lib/agent/index.ts          Orchestrator
app/api/agent/route.ts      API: POST (chat) + GET (memory/reset)
app/page.tsx                Dashboard
components/ChatPanel.tsx             Chat interface
components/MemoryInsightsPanel.tsx   Live memory sidebar
components/Header.tsx                Nav + wallet connect

Agent Flow

User Input → Fetch Memory → Extract Signals → Update Behavior Tags → Build Prompt → AI Inference → Handle Action → Persist → Respond

Memory Structure

Stored in .trademind-storage/<userId>.json (simulates 0G KV):

{
  "userId": "abc123",
  "risk_profile": "aggressive",
  "user_behavior": [{ "tag": "fomo_driven", "count": 3, "description": "..." }],
  "trade_history": [{ "asset": "ETH", "direction": "long", "simulated": true }],
  "notes": ["[4/22] fomo_driven detected: 'is it too late to buy ETH?'"],
  "interaction_count": 7
}

Detected Behavior Patterns

Tag	Trigger
late_entry_anxiety	"too late", "missed it", "already pumped"
fomo_driven	"FOMO", "everyone is buying", "mooning"
risk_averse	"safe", "too risky", "worried"
panic_seller	"sell all", "crashing", "dump"
high_conviction	"all in", "double down", "ape in"
methodical_accumulator	"DCA", "average in", "scale in"
bearish_tendency	"short", "overvalued", "bubble"
long_term_holder	"HODL", "long term", "patience"

API

POST /api/agent — { userId, message } → { message, reasoning, action, updated_memory, behavioral_signals }

GET /api/agent?userId=... — fetch memory

GET /api/agent?userId=...&action=reset — reset memory

Production Path (0G Integration)

1. Replace lib/storage/index.ts with 0G SDK (readKV, writeKV, appendLog)
2. Set OPENAI_BASE_URL to your 0G Compute node
3. Add real wallet signing via ethers.js (already installed)