NeuroPulseFlow
NinjaTrader-native, local-first AI-assisted trading software that connects risk-managed execution with transparent journals, quantitative review, web-grounded research, and operator-controlled strategy templates — without taking custody of client funds.
What's built, running, and proven today — not vaporware.
Completed
- Full C# NinjaTrader strategy (tens of thousands of lines) — live and executing trades.
- Python AI controller with CUDA GPU acceleration running locally.
- Docker Compose microservice stack (10+ containers) — web dashboard, news pipeline, event bus, portfolio publisher.
- Redis Streams event bus for real-time inter-service messaging.
- Automated chart snapshot pipeline + active-provider AI analysis (every 15 min during sessions).
- RSS/Atom news aggregation with ticker extraction, deduplication, and enrichment.
- AI Trade Scanner with pattern matching, star scoring, and rhythm detection.
- Volumetric / order-flow analysis module (VBTest) with depth map visualization.
- Auto-generated trade journal with entry/exit reasons, gate-block diagnostics, and risk metrics.
- AI operator chat with local/cloud model profiles, image input, connection testing, and saved or distilled sessions.
- Web-grounded AI research using self-hosted SearXNG with a DuckDuckGo fallback and clickable sources.
- One-click Trade Feedback combining journal statistics, strategy context, and matched chart snapshots for AI review.
- Quantitative performance reports covering expectancy, profit factor, drawdown, risk-adjusted metrics, and feature relationships.
- Performance-informed NinjaTrader template creation with real parameter validation, preserved base templates, and operator approval.
- n8n workflow automation (chart analysis, SEC/finreg feed ingestion, news enrichment).
- DeskContext API serving live aggregated state for NinjaTrader + web dashboard.
- LLM usage tracking, model/provider health checks, and live automation heartbeat monitoring.
Planned / In Progress
- Production installer + versioned releases + update pipeline.
- Voice command gateway (push-to-talk STT → trade intents → TTS feedback).
- Persistent semantic memory across journals, news, research, documents, and conversations.
- Broader institutional-data tools and measured model evaluation.
- Expanded broker integrations beyond NinjaTrader (phased).
- Health checks, diagnostics, and customer support tooling.
The AI chat is connected to the product workflow rather than presented as a generic chatbot. Operators can review trading evidence, request current research, inspect strategy templates, and keep useful conversations while retaining control over every configuration change.
Named provider profiles support local OpenAI-compatible models and optional cloud models. Built-in discovery, connection tests, and runtime smoke checks help verify a model before it becomes active.
A user-controlled Web Search mode adds current SearXNG results to the model's context, falls back to DuckDuckGo when needed, and presents clickable sources with the answer.
Operators can attach chart images, preserve full conversations, or save AI-distilled key points for later use. Token and latency metadata support cost-aware model selection.
These live product screens show how the AI workspace supports research and review, converts measured trading performance into a testable NinjaTrader configuration, and lets each operator choose models that fit their own hardware and privacy preferences.
The selected local model explains its role inside NeuroPulseFlow: synthesize macro context, technical flow, sentiment, news, and strategy evidence into structured decision support. The same interface makes its limits explicit: it does not place trades or promise outcomes; the operator remains responsible for execution.
Trade Feedback loads journal results, quantitative metrics, existing NinjaTrader templates, and real parameter values into one review. The AI identifies measured strengths and weaknesses, proposes a small set of parameter changes, and produces a new XML template without repetitive manual transcription. The operator still reviews, backtests, and explicitly selects the configuration before use.
The provider workspace discovers compatible local models, tests connectivity and runtime loading, saves reusable profiles, and exposes context controls for chat, search, and image reasoning. Users can select a lighter or more capable model based on their PC instead of being locked to one vendor or one hardware requirement.
NeuroPulseFlow closes the gap between execution and iteration with a human-controlled performance review loop.
- Capture: NinjaTrader writes structured trade records, context, outcomes, and gate diagnostics.
- Match: Available trades are paired with nearby chart-snapshot analysis for visual context.
- Measure: The metrics engine calculates expectancy, profit factor, drawdown, risk-adjusted results, R-multiples, and feature relationships.
- Review: One-click Trade Feedback asks the active AI model for a specific, quantitative assessment over a selected time window.
- Propose: The AI may recommend a small set of testable parameter adjustments and explain why.
- Validate and create: Suggested names are checked against the real NinjaTrader parameter catalog, then applied to a copy of an existing template.
- Approve and test: The operator reviews, simulates, and manually selects the new template; the original remains available for rollback.
These are actual NinjaTrader screenshots captured automatically by the NPF ChartSnap Exporter during live trading sessions. No mockups. No simulations. When timestamps align, chart analysis can be matched to journal trades and included in the structured AI review.
Rising Wedge detection, Bullish Divergence alerts, star scores (2.1–5.0), Pattern Match labels, rhythm analysis, support/resistance levels, and 5 sub-indicators (ADX, RSI, MACD, ATR, Order Flow Delta).
VBTest Data Collector showing volume profile, market depth, bid/ask imbalances, cumulative delta, and buy/sell volume breakdowns per bar — providing institutional-grade context.
Trading is not about never losing. It is about applying defined risk controls, recording outcomes consistently, and reviewing both good and bad decisions. These three sessions are real examples from the trade journal and can now feed the same structured performance-review workflow described above.
- A NinjaTrader 8 strategy + local Python integration that turns market inputs into trade decisions and risk-managed execution.
- Runs locally on the user’s machine (not purely server-side).
- Connects execution, journals, quantitative metrics, chart evidence, research, and controlled template iteration.
- Ships as templates + settings + documentation so the workflow is reproducible, inspectable, and reversible.
- Not a broker.
- Not a fund.
- Not a service that trades client money.
- Not a guarantee of profit.
- Tight integration with a mature trading platform (order management, charting, execution tools).
- Faster time-to-market vs building a full brokerage stack.
- Clear boundary: users trade their own accounts; we provide tooling.
- Multiple operating modes (Hybrid / ChartTrader-ATM oriented / AI-managed mode).
- Risk controls and safety governors; explicit stop/target configuration by mode.
- AI operator workspace with image input, saved sessions, optional web-grounded research, and local/cloud model choice.
- Trade Feedback and quantitative analytics for auditing performance over selectable time windows.
- Performance-informed template creation with parameter validation, manual approval, and rollback.
- Local-first architecture to minimize server dependency for core operation.
- Core C# strategy (NinjaTrader-native): real-time market handling, decision gating, order execution, and safety controls.
- Volumetric / order-flow module: additional liquidity-aware market context and validation workflows.
- Local Python + CUDA analysis: additional computation layer for analysis/inference while keeping the execution loop local.
- FastAPI + DeskContext: dashboard, live state aggregation, journal analytics, provider control, template operations, and health APIs.
- Redis + n8n: event and usage streams with scheduled news/chart workflows and observable heartbeats.
- Research layer: self-hosted SearXNG search with fallback routing and source-aware AI answers.
- Auto-generated trade journal/log artifacts for auditing and iteration.
- “Why it entered / why it didn’t”: decision snapshots, gate-block reasons, chart context, and measurable outcomes.
- Quantitative reports expose sample size, expectancy, drawdown, distributions, streaks, and feature relationships.
- AI-proposed settings are checked against real parameter names before a new template can be created.
- Operators approve and select templates manually, while preserved base templates provide a rollback path.
- Current artifacts: structured journals, AI review prompts, quantitative reports, chart snapshots, workflow health, and reviewable templates.
- Recorded live-trading video (OBS 1440p/60fps) with visible timestamps and trade logs, paired with the chart and journal evidence shown above.
- Planned standardized reporting: sample size, expectancy, drawdown, slippage assumptions, and “bad days.”
- Reproducible configuration: templates, documented lookbacks, and versioned settings.
- NinjaTrader Brokerage (tested baseline).
- For full liquidity/volumetric functionality, users may need paid market data + Level 2 data access (pricing varies; users verify in their account).
- Hardware stability matters (local AI + fast timeframes).
- Target price: $250 per month.
- The subscription lowers the upfront cost so the product is more accessible to middle-class retail traders while supporting recurring updates, security, model/workflow compatibility, and customer support.
- Initial year-one target: 1,000 active monthly subscribers, equivalent to $250,000 in monthly recurring revenue at full-price retention.
- Long-term scale objective: 10,000 sustained active monthly subscribers, equivalent to $2.5 million in monthly recurring revenue at full-price retention.
- Production hardening + delivery: installer, signed/versioned releases, update pipeline, security, stronger diagnostics, and customer support tooling.
- Measured intelligence + memory: persistent indexing across journals, research, news, documents, and conversations; broader institutional tools; model evaluation based on measurable usefulness.
- Operator experience: richer performance reports, controlled workflow automation, and reliable push-to-talk STT/TTS while keeping NinjaTrader as the focused execution platform.
- Product hardening (stability + support tooling).
- Distribution and customer acquisition.
- Compliance/legal + customer agreements.
- Production delivery, evidence publication, and faster iteration on roadmap features validated by demand.
- Capital + introductions (trading communities, distribution, enterprise relationships).
- Legal/compliance mentorship.
- Product + growth advisory.