Port of the same multi-grid MM bot to OndoPerps (api.ondoperps.xyz). Same ladder strategy, throttle, orphan tracking, exit-only / per-pair halts as the Lighter and Rise builds; new client layer for OndoPerps's HMAC-SHA256 auth scheme.

OndoPerpsMM/
├── accounts/ondo1/
│   ├── .env.template          # copy to .env, fill in ONDO_KEY_ID / ONDO_API_SECRET
│   ├── data/                  # CSV trade log
│   └── logs/                  # one logfile per run
├── bots/mm/ondo_mm/
│   ├── multi_grid_bot.py      # main bot — strategy ports 1:1
│   └── ondo_client.py         # HMAC-SHA256 signer + REST + WS helpers
├── common/
│   └── control_server.py      # bot's HTTP /status + /start + /stop + /close
├── dashboard/
│   ├── index.html             # single-page console (vanilla JS, no build)
│   └── serve.py               # tiny http server: python dashboard/serve.py
└── requirements.txt

The bot embeds a control server on CONTROL_PORT (default 8140). A single-page console at dashboard/index.html polls /status every 2s and offers Start / Stop / Cancel-orders / Close+flatten controls.

Normally you start it via start.py (above). If you want to run it standalone against a bot that's already running:

python dashboard\serve.py
# Then open http://127.0.0.1:8141  in your browser

If you point the dashboard at a remote bot, pass query params: http://127.0.0.1:8141?host=192.168.1.10&port=8140.

The dashboard shows:

• Equity, session PnL, realized + unrealized totals, net notional cap usage
• Per-pair table with position, mid, avg-entry, BBO spread (bp), realized session & all-time, halt reason — sortable by any column
• Recent fills tape
• Read-only config snapshot (grid params, caps, trading-armed flag)
• Halt banner when the bot trips a stop-loss, fill-imbalance, or session-DD guard

# 1. Create the venv
python -m venv venv
.\venv\Scripts\Activate.ps1
pip install -r requirements.txt

# 2. Provision .env
cp accounts\ondo1\.env.template accounts\ondo1\.env
# Edit: paste ONDO_KEY_ID (with ondoKeyId_ prefix) and ONDO_API_SECRET
# (with ondoApiSecret_ prefix) from the OndoPerps web UI.

# 3. One-command start (bot + dashboard + browser):
python start.py

start.py launches the bot and the dashboard static-server in their own process groups, opens http://127.0.0.1:8141 in your browser, and streams both processes' output to one terminal with colored prefixes:

[   bot    ] 2026-06-02 19:30:01,234 INFO config: pairs=...
[ dashboard ] Dashboard serving at http://127.0.0.1:8141

Ctrl+C in that terminal does a graceful shutdown — on Windows it sends CTRL_BREAK to the bot's process group so its asyncio finally block runs (which cancels resting orders if TRADING=true). Force-kill is a 15s safety net.

Useful flags:

python start.py --account ondo2          # different accounts/ subfolder
python start.py --no-browser             # skip auto-open
python start.py --dashboard-port 9000    # override dashboard port
python start.py --flatten-on-stop        # POST /close?flatten=true on Ctrl+C

OndoPerps lists 22 perpetuals (no crypto majors):

• Equities (16): AAPL, AMD, AMZN, COIN, CRCL, GOOGL, HOOD, INTC, META, MSFT, MSTR, NFLX, NVDA, ORCL, PLTR, TSLA (all -USD.P)
• Indices (2): US500-USD.P, US100-USD.P
• Commodities (3): XAU-USD.P, XAG-USD.P, WTI-USD.P
• ETF (1): DRAM-USD.P

Max position size per symbol per account: $500k. Leverage up to 10-20× depending on asset class.

Because OndoPerps charges 1.5 bp maker, the bot's grid step needs to clear at least ~3 bp per round-trip just to break even on fees:

realized per round-trip ≈ 2 × GRID_STEP_BP − (2 × 1.5 bp fee) − adverse_selection

A GRID_STEP_BP < 4 will lose to fees alone even before adverse selection. A reasonable starting config on this venue:

GRID_STEP_BP=15
LEVELS_PER_SIDE=1
ORDER_SIZE_USD=20
MAX_NOTIONAL_PER_PAIR_USD=100
REQUOTE_DRIFT_BP=20
PAIR_STOP_LOSS_USD=-1.5

All the same risk env vars as the Lighter/Rise bots (PAIRS_INCLUDE, ORDER_SIZE_USD, LEVELS_PER_SIDE, GRID_STEP_BP, MAX_NOTIONAL_PER_PAIR_USD, REQUOTE_DRIFT_BP, SESSION_DD_LIMIT_USD, PAIR_STOP_LOSS_USD, FILL_IMBALANCE_RATIO, etc.). Plus OndoPerps-specific:

• ONDO_REST_BASE_URL — https://api.ondoperps.xyz
• ONDO_WS_URL — wss://api.ondoperps.xyz/ws
• ONDO_KEY_ID — ondoKeyId_... (from the web UI)
• ONDO_API_SECRET — ondoApiSecret_...

These were discovered during initial wiring against api.ondoperps.xyz mainnet. Listed here so future ports / debugging skip the round-trip:

bots/mm/ondo_mm/_test_ws.py was the diagnostic tool that surfaced the auth-order issue; ONDO_DEBUG_WS=1 in .env enables raw-frame logging across all four WS consumers and was how the array-vs-object level encoding was caught.

• API key authentication
• Documentation index
• Create order · Cancel · Batch cancel
• Markets · Portfolio · Positions
• WS connect · WS login
• Channels: orders · positions · fills · depth book