Agent Configuration

Each agent is configured at creation time via the Hub API. This page covers all configuration options.

Creating an Agent

POST /api/agents
Authorization: Bearer <token>
Content-Type: application/json

{
  "symbol": "BTC/USDT",
  "strategy_name": "momentum_breakout",
  "interval_seconds": 1,
  "live_mode": false,
  "risk_params": {
    "max_position_usd": 1000,
    "stop_loss_pct": 2.0,
    "take_profit_pct": 4.0,
    "leverage": 1
  }
}

Configuration Fields

Risk Parameters

The risk_params JSONB field controls position sizing and risk management:

Dynamic TP/SL

When the LLM supervisor is enabled, it can dynamically adjust stop-loss and take-profit levels based on:

• ATR-based stop-loss calculation
• Automated risk:reward ratio
• Market condition changes

The supervisor runs on a configurable cycle (every N ticks on open positions).

LLM Configuration

The LLM provider is configured via secrets (environment variables or encrypted DB storage):

LLM Planning Cycle

1. Agent collects market data (candles, indicators, price history)
2. LLM analyzes the data and selects the best strategy
3. LLM defines lookback period, threshold, and max loss percentage
4. Strategy runs on the selected parameters
5. LLM reasoning is captured for audit trail

Execution Modes

Paper Trading (Default)

• live_mode: false
• Positions tracked in-memory
• No real orders placed
• Full P&L simulation

Live Trading

• live_mode: true
• Requires HASHKEY_API_KEY and HASHKEY_SECRET
• Orders placed on HashKey Global exchange
• Set HASHKEY_SANDBOX: true for sandbox testing

Per-Agent Secrets

Override user-level secrets for specific agents:

POST /api/agents/{id}/secrets
Authorization: Bearer <token>
Content-Type: application/json

{
  "key_name": "OPENAI_API_KEY",
  "encrypted_value": "<AES-encrypted-value>"
}

Secret resolution order (first match wins):

1. Ephemeral override (per-request, never stored)
2. Agent-specific override (encrypted in DB)
3. User-level secret (encrypted in DB)
4. Process .env fallback

Agent Lifecycle

Management Commands

# List all agents
GET /api/agents

# Get agent details
GET /api/agents/{id}

# Stop an agent
POST /api/agents/{id}/stop

# Delete an agent (removes container)
DELETE /api/agents/{id}

Log Levels

Agent logs use structured levels for filtering: