d45dde8458
20-commit atomic ladder (X0–X19) + Phase 1→2 gate + Phase 2 runtime
runbook. Implements spec da1dd92bf:
X0: perception_forward_golden fixture (bit-equivalence gate)
X1: CfcTrunk v2 weight skeleton (no callers)
X2–X9: incremental weight-group migrations (VSN, Mamba2 ×2, LN ×2,
attn-pool, CfC, GRN heads), each gated by golden fixture
X10: hoist forward kernels into CfcTrunk methods
X11: capture_graph_a covers full v2 forward + captured-vs-uncaptured
equivalence test
X12: CheckpointV2 envelope + save/load (V1 hard-rejected)
X13: PerceptionTrainer.save_checkpoint delegate
X14: alpha_train saves best_h6000 checkpoint
X15: verify ml-backtesting accepts CheckpointV2 (no code change)
X16: max_drawdown_pct with \$35k base
X17: emit_deployability_verdict + tiered logic + 6 unit tests
X18: GPU smoke test against real trained checkpoint
X19: three sweep YAMLs (smoke, threshold-tuning, deployability)
Gate: fold-0 smoke must reproduce recorded 3-fold A/B numbers
within ±0.010 absolute before Phase 2 begins
P.1–6: Argo runtime (training → smoke → threshold → sweep → verdict)
Self-review confirms 1:1 spec coverage. Three soft adaptation points
(HEAD_MID constant, Mamba2Block accessors, BacktestHarnessConfig field
names) resolve at code-read time. One placeholder (todo!() in X11
explanatory text) is called out in self-review for replacement when
that commit lands.
Co-Authored-By: Claude Opus 4.7 (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.