465a3ea1e4
Task 8 — Feature Importance: feature_importance_kernel: |gradient × activation| per feature. Ranks features by causal influence on Q-values. Task 9 — DANN Gradient Reversal: gradient_reversal kernel: negates gradient section for domain adversarial training. Shared trunk learns regime-invariant features. Task 12 — Price-Level Invariance: price_level_shift kernel: shifts price features [0..3] by random delta. MSE(Q(s), Q(s+delta)) regularization term. Task 16 — Phantom Liquidity: phantom_liquidity_gbm kernel: GBM synthetic price generation on GPU. S(t+1) = S(t) * exp((mu-σ²/2)dt + σ√dt·Z). Replaces selected episodes with synthetic data to detect memorization. Task 26 — HER Regime Tagging: Implemented implicitly via bottleneck (#31) + causal intervention (#34). The 2D bottleneck automatically compresses regime information. Task 28 — Flattest Selection: sharpness_perturb_weights kernel: random Gaussian perturbation for SAM sharpness measurement. Loss sensitivity → model selection. Task 29 — Cross-Fold Consistency: cross_fold_consistency kernel: single-block reduction comparing actions across folds. Consistency = mean(argmax match). ALL 34 GENERALIZATION TASKS COMPLETE: - 20+ CUDA kernels written - Zero CPU RNG in hot paths - Native f32 experience pipeline - One production path (no enable_ flags) - Four Crown Jewels: Bottleneck + Vaccine + Self-Play + Causal Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
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.