Current status — October 20, 2026: the DigitalOcean trading server was permanently removed on September 14, 2026. The Cloudflare terminal is separate from the removed trading API and daemon. Runtime diagrams, accounts, performance figures, and operator commands below are historical. See current hosting and retirement status.
QGTM GOLD/SILVER SYSTEM -- INTELLIGENCE BRIEF
Historical generated brief. It is timestamped and not the live operational source of truth. Use
/api/v1/strategies,/api/v1/session/readiness, and the deployment/runbook docs for current production state.
Generated: 2026-04-12 | Status: OPERATIONAL (Paper Mode) | IVRA Verified
SYSTEM HEALTH
| Component | Status | Detail |
|---|---|---|
| Platform | API, Terminal, CI, WebSocket all live | qgtmai.com + api.qgtmai.com |
| Tests | 3,077 passing | 90.71% coverage, lint clean |
| Gates | 11/15 GREEN | Open: 4 (all require live infra or API keys) |
| Kill-switch | ARMED (untested) | 5-tier with 2-person reset |
| Regime | TRANSITION (default) | HMM+BOCPD ensemble not fed live data yet |
| Gaps | 9 open (17/26 resolved) | 4 SEV1, 4 SEV2, 1 SEV3 -- all infrastructure |
EDGE ARCHITECTURE (38 strategies across 9 categories)
| Category | Count | Status | Expected Sharpe | Key Dependencies |
|---|---|---|---|---|
| Macro Regime | 5 | 5 active | 0.3--0.8 | FRED, DXY, TIP |
| Curve & Carry | 5 | 2 active, 3 BLOCKED | 0.4--0.9 | Futures data (GC/SI/MGC) |
| Relative Value | 6 | 6 active | 0.5--1.0 | ETF prices only |
| Positioning & Flow | 6 | 5 active, 1 BLOCKED | 0.3--0.7 | CFTC COT, COMEX/SGE data |
| Volatility | 5 | 3 active, 2 shadow | 0.2--0.6 | Options chain data |
| Event & Seasonal | 3 | 3 active | 0.2--0.5 | Intraday data |
| ML / Ensemble | 3 | 3 shadow | 0.4--0.8 | Training data, PBO gate |
| Systematic | 4 | 4 active | 0.5--1.2 | Price history |
| Structural | 1 | 1 active | 0.3--0.7 | Multi-source |
| TOTAL | 38 | 29 active, 5 shadow, 4 blocked |
Blocked strategies: 4 curve/carry + positioning (no futures/CFTC feed).
CAPITAL & INCOME PROJECTION
Formula: Monthly = Capital x DSR x Vol_Target x (1/sqrt(12)) - Costs
95% CI: +/- 1.96 x Capital x Vol_Target x (1/sqrt(12)) / sqrt(12)
Assumptions: DSR = 1.50, Vol Target = 15%, Monthly Costs = $500
| Capital | Monthly Income (est.) | 95% CI | Annual |
|---|---|---|---|
| $100,000 | $5,995/mo | $3,545 -- $8,445 | $71,942 |
| $250,000 | $15,738/mo | $9,613 -- $21,863 | $188,856 |
| $500,000 | $31,976/mo | $19,726 -- $44,226 | $383,711 |
| $1,000,000 | $64,452/mo | $39,952 -- $88,952 | $773,423 |
| $3,000,000 | $194,356/mo | $120,856 -- $267,856 | $2,332,269 |
| $6,000,000 | $389,211/mo | $242,211 -- $536,211 | $4,670,537 |
\(100K/mo target requires ~\)3--6M at sustained edge.
Honest note: DSR of 1.5 is top-decile. Median systematic PM fund achieves 0.8--1.2. Until backtests are re-run post-bug-fix, all performance claims are theoretical. The 3 critical bugs fixed (lookahead, purged CV, OOS fabrication) mean prior numbers are invalidated.
WHAT WORKS RIGHT NOW
- Paper trading via Alpaca (GLD, SLV, GDX, IAU, SGOL, SIL, PPLT, UGL, AGQ, +35 more)
- Live
/api/v1/strategiesreturned 31 live-known strategies on 2026-04-16, with 26 active and 5 inactive/quarantined at check time - Signal aggregation: inverse-vol x skill x capacity x meta-label weighting
- Regime-conditional allocation (risk-on / risk-off / crisis / inflation / transition)
- Kill-switch (5-tier), reconciliation (60s), watchdog (120s timeout), Merkle audit log
- Self-learning loop: decay detection -> retrain -> shadow -> promote pipeline planned (modules built + tested but not wired to scheduler — see
scripts/dead_ml_loop/README.md). Live decay scoring runs viaqgtm_risk/decay_monitor.py(/api/v1/risk/decay/*REST). - 6 execution algos (VWAP, TWAP, IS, Iceberg, Sniper, POV) with TCA
- 9-factor Barra-style risk model + EVT tail risk + 10 stress scenarios
- 47 REST routes + WebSocket terminal + Prometheus /metrics endpoint
WHAT'S BLOCKED (owner action items)
| # | Action | Unlocks | Effort |
|---|---|---|---|
| 1 | Set Alpaca API key in env | Backtests, PBO/DSR validation, live paper signals | 2 min |
| 2 | Run scripts/run_backtest.py |
Quarantine strategies failing PBO < 0.5 | 30 min |
| 3 | mkdocs gh-deploy |
Published docs site (87 pages, builds clean) | 1 min |
| 4 | Deploy Grafana + connect OTel | Latency SLO measurement (p99 < 50ms gate) | 2 hr |
| 5 | Connect IBKR paper account | Second broker redundancy, FIX protocol | 1 hr |
| 6 | Fire-drill kill-switch in staging | Validates 5-tier emergency shutdown | 30 min |
| 7 | Wire CFTC COT + COMEX feeds | Unblocks 4 positioning/flow strategies | 2 hr |
NEAREST RISKS
| # | Risk | Severity | Mitigation |
|---|---|---|---|
| 1 | No live backtest results -- all performance is theoretical | SEV1 | Run backtests with corrected engine (item #1-2 above) |
| 2 | Single broker (Alpaca) -- no redundancy if outage | SEV1 | Connect IBKR paper (item #5) |
| 3 | Single droplet -- no HA failover | SEV2 | K8s manifests ready, need k3s cluster |
| 4 | Futures/FX data not connected -- 4 strategies blind | SEV2 | Wire data feeds (item #7) |
| 5 | Kill-switch never fire-drilled | SEV1 | Stage drill with simulated positions (item #6) |
| 6 | Audit log Merkle chain not independently verified in prod | SEV2 | Run scripts/verify_audit_chain.py on live data |
COMPETITIVE POSITION
Built: 38 strategies, 9-factor PM risk model, self-learning pipeline, Merkle audit log, compliance framework, 18 SR 11-7 inventoried models, 6 execution algos, EVT tail risk, 10 stress scenarios, 3,077 tests at 90.71% coverage.
Comparable to: Mid-tier systematic precious metals fund infrastructure. Code quality and test coverage exceed many production hedge fund codebases.
Gap to production: Live track record (no substitute), futures execution (ETF-only today), institutional broker breadth (Alpaca only), deployed observability (OTel wired but Grafana not live), HA deployment (single node).
NEXT 30 DAYS (priority order)
| Week | Actions | Outcome |
|---|---|---|
| Week 1 | Set API key, run backtests, quarantine PBO failures | Validated strategy suite with honest DSR |
| Week 2 | Publish docs, deploy Grafana, connect IBKR paper | Full observability + broker redundancy |
| Week 3 | First live paper trading with full validated suite | Real signals hitting real (paper) market |
| Week 4 | Review 30-day paper results, compute live DSR | First honest performance number |
30-day success metric: Live paper DSR > 0.8 across the validated strategy subset.