0275a25d9f
SP11 reward-as-controlled-subsystem: B1b z-score normalization for
magnitude-asymmetric weight ratios; popart-component magnitude slot.
SP12 v3: per-trade event-driven reward (asymmetric pos/neg cap +
min-hold target + zero per-bar dense shaping).
SP13 Layer B: K=1→2 softmax CE aux head + i32 replay buffer ring +
GpuBatchPtrs plumbing + scale-free MSE bridge for ISV[117].
SP14 Layer A+B: 3 stability fixes (C51 inv_a_std floor lift, set_aux_weight
clamp, stagnation warmup gate) + Earned Gradient Flow pearl wired through
alpha_grad_compute / q_disagreement_update / gradient_hack_detect /
dir_concat_qaux / scale_wire_col kernels + ISV_TOTAL_DIM bus fix
(383→396) + warmup_gate delete (variance-driven k_aux/k_q handles
warmup intrinsically).
Smoke A2-B PASSED. 8-epoch train-dd4xl L40S validation revealed:
- Walk-forward test_start..test_end slice generated but never consumed
(val IS the selection set; no sealed test).
- Downward-spiral pathology: trades 131k→64k, active_frac 0.48→0.17,
sharpe_ann 79→44 across 8 epochs.
SP15 (trader-discipline-and-recovery) addresses both: honest cost-aware
metrics on Q1-Q7/Q8/Q9 split, behavioral test suite on dev RTX 3050 Ti,
DD-state foundational input, recovery dynamics inc. plasticity injection.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
# Conflicts:
# docs/superpowers/specs/2026-05-04-sp11-reward-as-controlled-subsystem.md
Foxhunt
Production HFT trading system in Rust.
Architecture
The workspace contains 32 crates organized as follows:
Core Libraries (16)
| Crate | Purpose |
|---|---|
trading_engine |
Order processing, FIX 4.4, IB TWS, SIMD, RDTSC timing |
risk |
VaR, Kelly, circuit breakers, kill switches, compliance |
risk-data |
Risk data types and shared structures |
trading-data |
Trading data types |
ml |
DQN Rainbow, PPO, TFT, Mamba2, ensemble inference |
ml-data |
ML data types and feature definitions |
data |
Market data ingestion and storage |
backtesting |
Replay engine, strategy tester |
adaptive-strategy |
Ensemble execution, microstructure analysis |
common |
Shared types, resilience, error handling |
storage |
S3 and local model storage |
model_loader |
Model serialization and loading |
market-data |
Market data feed handlers |
database |
PostgreSQL access layer (SQLx) |
config |
Configuration management |
tli |
CLI commands and tooling |
Services (8)
| Service | Purpose |
|---|---|
backtesting_service |
gRPC backtesting service |
broker_gateway_service |
FIX routing, broker connectivity |
trading_service |
Core trading operations |
ml_training_service |
Model training orchestration |
data_acquisition_service |
Market data acquisition |
trading_agent_service |
Autonomous trading agents |
api_gateway |
gRPC API gateway with auth |
web-gateway |
Axum REST + WebSocket gateway |
Frontend
web-dashboard/ -- React 19 + TypeScript + Vite + TradingView charts.
Building
# Check compilation (no PostgreSQL required)
SQLX_OFFLINE=true cargo check --workspace
# Run tests for a specific crate
SQLX_OFFLINE=true cargo test -p <crate> --lib
# Clippy
SQLX_OFFLINE=true cargo clippy --workspace
ML Models
Four production model architectures on Candle v0.9.1 with CUDA:
- DQN Rainbow -- Deep Q-Network with prioritized replay, dueling heads, noisy nets
- PPO -- Proximal Policy Optimization with GAE and LSTM policies
- TFT -- Temporal Fusion Transformer for multi-horizon forecasting
- Mamba2 -- State space model for sequence prediction
Each model has a standalone trainer and a UnifiedTrainable adapter for the hyperopt pipeline.
Infrastructure
- Git: Gitea at
git.fxhnt.ai(Tailscale-only), Scaleway DEV1-S - Observability: OpenTelemetry OTLP (env
OTEL_EXPORTER_OTLP_ENDPOINT) - Database: PostgreSQL with SQLx offline mode for CI
License
Proprietary. All rights reserved.