MM 3-Layer Stack

Sequence MM is now organized into three public layers so clients can build with normal Rust tooling and no central-coordinator internals.

Layer Model

Template crates:

• mm-algo-template: WASM MM strategy starter
• mm-agent-template: Native MM agent starter

What Changed

Before:

• MM config types were duplicated across strategy and agent code
• API payloads used weakly typed config values
• Edge MM control handlers were partial/stubbed

Now:

• One canonical MM schema layer (mm-types)
• Typed client methods for deploy/adjust/halt/status/agent endpoints
• Working runtime config injection and heartbeat-driven status

Canonical Types

MmConfig

MmConfig is the canonical strategy config:

• #[repr(C)], 56-byte layout for WASM boundary
• JSON serde for REST use
• Alias normalization for legacy names:
  • spread_bps -> min_spread_bps
  • size_1e8 -> quote_size_1e8
  • order_ttl_ms -> max_order_age_ms
  • skew_factor -> max_skew_ratio

Validation rules:

• num_levels in 1..=5
• min_spread_bps > 0
• quote_size_1e8 > 0
• max_position_1e8 > 0

MmRiskLimits

Canonical risk limits shared by REST control and edge deploy:

• max_position_1e8 > 0
• max_order_notional_1e9 > 0
• max_order_rate > 0

MmHeartbeat

Unified heartbeat model for status and regime detection, including enriched fields like:

• realized_volatility_bps_1min
• trend_bps_1min
• top_level_spread_bps
• quote_fill_ratio_bps

Client Flow

Typical deployment/control flow:

1. Build strategy from mm-algo-template to WASM.
2. Deploy via MmControlClient::deploy(...) with typed config/risk.
3. Monitor via status() and features().
4. Adjust in place via adjust_config(...).
5. Halt with halt(...) if needed.

Minimal usage:

use mm_control_client::MmControlClient;
use mm_types::{MmConfig, MmRiskLimits};
 
let cc = MmControlClient::new("https://api.sequencemkts.com", "seq_live_...");
 
let cfg = MmConfig::default();
let risk = MmRiskLimits::default();
 
// Deploy
let wasm_bytes: Vec<u8> = std::fs::read("algo_btc_usd.wasm")?;
let deployed = cc.deploy("kraken", "BTC-USD", &wasm_bytes, cfg, risk).await?;
 
// Adjust config
let mut next = cfg;
next.version += 1;
cc.adjust_config(&deployed.algo_id, next).await?;

Migration Guide

Current status:

• mm-algo-core and mm-agent remain available for transition
• New development should target template crates

Runtime Behavior Changes

The MM runtime path now enforces stricter behavior:

• Invalid/malformed order types are rejected rather than guessed
• Risk-rejected actions are neutralized before OMS submission
• MM status config_version reflects injected config version from runtime