Oscillators
momentum & strength —| Indicator | Value | Action |
|---|
Moving Averages
trend & crossover —| Indicator | Value | Action |
|---|
Pivot Levels
classic · fibonacci · camarilla · woodie · demark| Pivot | Classic | Fibonacci | Camarilla | Woodie | DM |
|---|
Real-Time Candlestick & Volatility Visualizer
Market Microstructure
Interactive Live Simulation Ready
Inject mock ticks to test the model's live response. Simulate random walk, trending structure, or deep swing sweeps.
Replay Mode & Accuracy Tracking
Load historical market snapshots and play them through the system to analyze signal precision and stop triggers.
Historical Replay Session (ENGINE-BACKED) idle
Runs the REAL decision pipeline (causal features → 70D → model → policy → risk → simulated fills) over local historical data through a logical replay clock. The chart cursor separates KNOWN from UNKNOWN — future candles never enter decision state.
System Health & Diagnostics
Database Health & Hygiene
Risk Plan (RiskEngine — single source of truth)
—Active Open Positions & Execution modification
0 Open| TICKET | SYMBOL | DIRECTION | LOTS | OPEN PRICE | STOP LOSS | TAKE PROFIT | PROFIT (USD) | ACTIONS |
|---|---|---|---|---|---|---|---|---|
| No active positions currently registered on broker. | ||||||||
Closed Trading History
| ORDER ID | SYMBOL | LOTS | EXECUTION | PRICE | PROFIT / LOSS | CLOSED TIME |
|---|---|---|---|---|---|---|
| Loading closed trade history… | ||||||
Account Performance & Intelligence
| STRATEGY | TRADES | NET PnL | WIN RATE | PROFIT FACTOR | AVG R | LIFECYCLE | CONFIDENCE |
|---|---|---|---|---|---|---|---|
| NO STRATEGY EVIDENCE AVAILABLE | |||||||
--
Feature Category
Volatility & Microstructure Features (50-Dimensional Matrix)
Normalized Log Scale [-3.0, +3.0]PyTorch ScalpNet v3 Reasoning (TCN + Attention)
Neural activation output vectors representing probability distribution across multi-confluence trading scenarios.
AI Prediction vs Actual Market Movement
| TIME | PRED ACTION | CONFIDENCE | ACTUAL DELTA | OUTCOME |
|---|---|---|---|---|
| No recent AI prediction outcomes evaluated yet. | ||||
Trade Intelligence Brain
worker: --Market Radar
NO RADAR DATAStrategy Research & Validation Engine
EVIDENCE-DRIVEN GATES: discovery → backtest → walk-forward → OOS → robustness → score. A candidate NEVER goes live automatically.
Strategy Research Detail
One-click trace: strategy → runs → gates → events → evidence → snapshot. Shows exactly why a strategy is waiting.
Validate a Candidate
Runs the full gate chain (backtest → walk-forward → OOS → robustness → score). Result is VALIDATED or REJECTED — never ACTIVE.
Strategy Factory — Autonomous Strategy Evolution
GENERATE → VALIDATE → BACKTEST → WALK-FORWARD → OOS → ROBUSTNESS → SCORE → RANK → ELITE → EVOLVE. Candidates NEVER go live automatically.
Set an OpenAI-compatible endpoint + API key (e.g. http://178.105.20.69:20128/v1). When configured, each generation is assisted by your local AI model; otherwise deterministic generators are used.
Factory Event Stream
Structured lifecycle events (persisted, immutable).
Factory Console — live endpoint results
Every generate/loop/save call result, error and warning lands here (newest at bottom). Debug without the browser devtools.
Failure Analysis
Rankings (overall)
Strategy Benchmarks — per-candidate filtered backtest
Each candidate is graded on its OWN filtered ledger slice (DSL-aware replay). Not the ledger average — scores diverge so the AI can rank. Walk-forward / OOS / robustness run on the filtered subset.
News Intelligence Engine
PERSISTENT NEWS INTELLIGENCE: canonical events, source consensus, XAUUSD/USD impact, time decay. News can never force a trade.
Advanced Price-Hunting & Scalping Rule Matrix
PERSISTENCE CONTROL: All 30+ institutional rules are disabled by default. Enabling a rule overrides standard PyTorch AI.
Algorithm Live Tuner
Tweak quant algorithm thresholds on the fly without restarting the engine.
Multiplier for ATR stop-loss buffer (0.5 to 4.0).
Minimum required Risk-to-Reward ratio (1.0 to 5.0).
Min Neural Net score to validate liquidity boxes (0.50 to 0.99).
Penetration depth before invalidating an FVG (0.1 to 1.0).
Historical bar depth for Order Block scanning (10 to 100).
Configuration Engine (live.yaml)
Directly edit bot configurations. All values update on disk dynamically.
DEBUG & DIAGNOSTICS HUB — 70D RUNTIME INTELLIGENCE
One canonical snapshot of the entire active pipeline: raw market state, feature calculation, 70D vector, scaler, model input/output, policy, risk, exposure, execution, liquidity, news, workers, database, caches, chart and SSE. The UI only renders backend state — it never computes its own trading intelligence.
RUNTIME STATUS
Real backend state — never fabricated defaults.
Subsystem Health
70D FEATURE MATRIX
Registry-driven rows (schema_contract). RAW / NORMALIZED / CLIPPED shown only when the runtime exposes that stage; otherwise NOT_EXPOSED — never a fake zero.
| Idx | Name | Family | RAW | NORMALIZED | CLIPPED | FINAL | Status | Source | Causality |
|---|---|---|---|---|---|---|---|---|---|
| Awaiting feature stream... | |||||||||
MODEL INPUT / OUTPUT
The exact tensor the live path consumed and the model output it produced.
CONFIDENCE PIPELINE
Every real transformation the runtime applied. Stages that do not exist are not shown.
POLICY DECISION TRACE — WHY THIS TRADE / WHY NO TRADE
Every gate in order with actual value / threshold / status.
RISK ENGINE
Canonical account + risk limits + the risk engine decision.
EXPOSURE — INTERNAL vs BROKER TRUTH
Detects stale-cache bugs (internal cache vs broker truth).
EXECUTION STATE
Order manager state + broker connection.
POSITION MANAGEMENT
Open positions with MFE / MAE / protection state.
EXIT FORENSICS
Why a position stayed open or was not closed.
MARKET INTELLIGENCE ENGINE — MARKET PERCEPTION LAYER
Regime, swing structure, liquidity map, sweep events, breakout quality, smart-money features. Pure perception — the final decision stays with the strategy models.
LIQUIDITY INTELLIGENCE — 10D (60..69) + CONTEXT
Why bsl_distance_atr = X: active pools, nearest BSL/SSL, EQH/EQL, HTF, confluence, sweep state.
NEWS INTELLIGENCE
Canonical news context + which news dimensions are active in the model (50..59).
WORKER STATUS
A worker marked RUNNING but idle too long is flagged DEGRADED.
DATABASE STATE
Path / size / WAL / health (backend snapshot — no per-request DB scans).
SYSTEM CACHE STATE
Model / feature / liquidity / news / exposure / chart / research caches.
CHART DEBUG + SSE
Chart data source / bars / overlays + SSE connection health.
DEBUG SNAPSHOT
Capture the current snapshot, compare two snapshots (T0/T1 feature diff), or expand the raw JSON tree.
MT5 IPC Telemetry Console
Broker execution retcodes, IPC latency and order state transitions, newest first.
Model Governance & Champion/Challenger
Truthful registry state, load gate, shadow runtime, feature/news parity, latency, promotion lifecycle.
Champion
--Challenger
--Shadow Runtime
Latency (ms)
Promotion State
Recent Governance Events
Promotion Controls (70D)
70D Shadow Model
scalp_v3 (50 Base + 10 News + 10 Liquidity). Observability only — no execution authority. Live feature health, drift, disagreements.
Model
Runtime
Disagreement
Liquidity Feature Health (LIVE)
Recent Disagreements
Liquidity Intelligence
70D integration contract: Base 0..49 | Family 50..59 | Liquidity 60..69. Structural information only — never a trade signal.
Forensic Incident Center
OPEN INCIDENTS
--
CRITICAL
--
HIGH
--
MEDIUM
--
INCIDENT WORKER HEALTH
DISABLEDno worker attached
One-Click Trace (spec 37)
Enter a ticket / execution_id / request_id / model_id / incident_id to trace its full lineage.
Incidents
DATABASE MANAGEMENT
Current Database
Provider Switch
Switching the provider does NOT move data. Run the migration workflow first, then switch.
PostgreSQL Configuration
Migration Workflow (SQLite → PostgreSQL)
Database Explorer (all database files — auto-syncs when new DBs are added)
NEURAL MODEL STUDIO & DEEP LEARNING
50DFull-transparency neural network inspection, interactive prediction, 70D live assembly, and dataset training. Every artifact written below reports its real on-disk location.
ARTIFACT LOCATOR
RESOLVING70D MULTI-SOURCE COMPONENT ASSEMBLY
WAITING FETCHAssembles 50 Base features (live tick stream) + 10 News features (NewsEngine) + 10 Liquidity features (LiquidityGovernor) under strict schema hash verification.
INFERENCE CONTROLS
PROBABILITY DISTRIBUTION & DECISION
LAYER-BY-LAYER NEURAL ACTIVATIONS & NORMS
Monitors forward hook activation L2 norms, means, and sparsity| Layer Name | Type | Tensor Shape | L2 Norm | Mean | Std Dev | Sparsity (Zero %) |
|---|---|---|---|---|---|---|
| Run neural inference to inspect internal activations. | ||||||
AI HUB: MODEL REGISTRY & RUNTIME HOT-LOADER
CHAMPION: NONEDynamically hot-load neural model weights and calibrated scaler sidecars into live engine memory without server restart. Supports persistent SQLite tracking, pre-load verification, fine-tune gating, and 1-click rollback.
--
--
| Check Name | Passed | Diagnostic Detail |
|---|
| # Index | Mean (μ) | Std Dev (σ) | Clamping | Zero Variance |
|---|
DATASET DOWNLOAD & HISTORICAL MARKET INGESTION
API FirstDownload and ingest historical market candles (1m, 3m, 5m, 15m) with automatic schema verification, zero-variance clamping, and monotonic UTC validation.
--
MODEL TRAINING DISPATCH & FEATURE INSPECTION
Algorithm Cycle ReadySelect a training dataset, inspect/normalize 50D & 70D features, and execute PyTorch model training.
| # | Family | Feature Name | Raw Mean | Raw Std | Norm Sample | Status |
|---|
LAYER-2 ML: POSITION MANAGEMENT DATASET GENERATOR
Risk & Continuation ModelSimulate trade decisions across historical bars to generate state samples (unrealized R, position age, ATR) with anti-leakage mathematical labeling (KEEP / CLOSE / REDUCE) and chronological purge/embargo splitting.
LAYER-2 ML: POSITION DECISION ADVISER
DISABLEDDeep-learning adviser for the decide system: on every live position it scores KEEP / CLOSE / REDUCE from real position state (unrealized R, age, ATR, spread). It can only ever make a close verdict more conservative — it never opens, sizes, or extends a position. Default is DISABLED, so the decide system runs exactly as before.
ADVERSARIAL STRESS TEST
Injects constant zero-variance features, flash-crash shocks (+-10^6), NaNs, and boundary dimensions.
| Scenario | Verdict | Detail |
|---|
LATENCY BENCHMARK
Measures forward pass latency percentiles and throughput on current hardware.