b4aadff756
Phase 2.0 seeded the V envelope at ±200 reasoning fromdd049d9a4dollar magnitudes ($211 avg win, $235 avg loss). But the envelope clamps the POST-SCALE V regression target — after apply_reward_scale divides by mean_abs_pnl_ema to keep typical reward magnitudes ~1.0, the target's natural range is `r + γ × v_tp1 ≤ ~3` in atom-support units, NOT dollars. A ±200 bootstrap was a no-op (envelope wider than any possible target excursion) until the EMA tightened it. ±10 is the tighter bootstrap matching the actual post-scale operating range while still wide enough to admit normal target excursions. The kernel's existing sentinel-zero bootstrap discipline at line 98 (pearl_first_observation_bootstrap) snaps the EMA to the first observed trade magnitude as soon as a done event fires, so this bootstrap only matters for the first ~10-20 steps. Note: this commit does NOT resolve the step-4 NaN that surfaced after the atomicAdd determinization (10d4614fb) — l_v is identical 6.329 ± 0.001 across all ±10 / ±200 / pre-fix runs, proving V predictions are not the proximate NaN cause. The tightening is still correct per codebase discipline (pearl_controller_anchors_isv_driven + feedback_isv_for_adaptive_bounds) and removes one masked confound for the next debug pass. Suspected actual NaN cause shifted to advantage RMS / encoder h_t / shared streaming statistics — see pearl_atomicadd_masks_v_instability for the open investigation. Co-Authored-By: Claude Opus 4.7 <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.