AlphaFusionNet — LLM‑Guided Neural + Graph Market Monitoring Engine

This repository is designed for market monitoring, risk awareness, market structure analysis, and data/ops governance.
It is not an investment advisory product and must not be used to generate or present:

• buy/sell or long/short instructions, entry/exit, target prices, TP/SL
• allocations, portfolio/weight recommendations, “best opportunities,” expected returns
• PnL, outperformance/underperformance claims, or performance marketing outputs

If you build UI/outputs on top of this repo, keep the language descriptive/diagnostic, not directive.

---

What this repo does

This repository provides an end‑to‑end pipeline that integrates four core components:

1. NeuralFusionCore
   Ingests recent OHLCV + news and produces per‑symbol model salience signals (internally stored as signed values, but can be exposed in compliance‑safe form as magnitude / tiers only).

2. ChronoBridge
   Generates time‑aligned fused embeddings (price + news) and persists them for downstream retrieval.

3. NetWeaver
   Learns cross‑symbol dependencies via a graph model and produces:

   • a rank/score per symbol (use as relative attention, not “expected return”)
   • an optional dependency graph describing probabilistic linkages.

4. AlphaFusionNet (Fusion + Reasoning Service)
   Fuses model signals and produces monitoring‑oriented outputs, such as:

   • risk/attention watchlists (top‑K symbols to monitor)
   • market/network structure summaries
   • governance artifacts (policy JSON, audit logs, data freshness flags)
   • diagnostic text (LLM‑generated), strictly constrained to risk/structure/monitoring language.

---

High‑level capabilities

• Real‑time multimodal ingestion (OHLCV + news) and feature synchronization
• Fused embeddings for downstream graph modeling
• Network dependency modeling (influence, lead‑lag linkages, clusters)
• Compliance‑safe dashboards:
  • Market snapshot (risk temperature, news heat, correlation stress, regime shift likelihood)
  • Risk & attention watchlist (symbols that require monitoring)
  • Risk pack table for the universe (volatility tiers, liquidity stress tiers, drawdown/tail risk buckets, data quality flags)
  • Network view (influence leaderboard, probabilistic influence map, contagion stress)

---

Table of contents

1. Architecture diagram
2. Required downloads
3. Setup
4. Pipeline
   • Run ChronoBridge API service
   • Run AlphaFusionNet API service
   • Run scheduler
   • Run tests
5. Script cheat‑sheet
6. Dependencies
7. Appendix
8. Authors & citation
9. Support

Setup

# Clone repository
git clone https://github.com/Novoxpert/AlphaFusionNet.git
cd AlphaFusionNet
git submodule update --init --recursive

# (optional) create a virtual environment
python -m venv .venv

# Linux/macOS:
source .venv/bin/activate

# Windows (PowerShell):
.\.venv\Scripts\Activate.ps1

# Install dependencies
# Linux:
pip install -r requirements.txt
# Windows:
pip install -r requirements-windows.txt

---

Pipeline

This pipeline coordinates ChronoBridge data services, model APIs, and scheduled tasks to produce synchronized features and monitoring outputs.
Full usage notes: ./docs/usage.md

1) Run ChronoBridge API service

API for retrieving fused embeddings and OHLCV per symbol from MongoDB.

python -m apps.ChronoBridge.scripts.chronobridge_api_service

2) Run AlphaFusionNet API service

API for retrieving latest monitoring outputs from MongoDB.

python -m scripts.alphafusionnet_api_service

3) Run scheduler

# Linux:
celery -A scheduler.tasks worker --loglevel=info --concurrency=1

# Windows:
celery -A scheduler.tasks worker --loglevel=info -P threads --concurrency=1

python run_triggers.py
celery -A scheduler.scheduler beat --loglevel=info

Typical outputs include processed parquet files, model checkpoints, and MongoDB collections such as: chrono_bridge, NeuralFusionCore_predictions, NetWeaver_predictions, AlphaFusionNet_predictions, live_metrics, monthly, windows, etc.

4) Run tests

pip install pytest
pytest -v tests

---

Script cheat‑sheet

For the full engine layout, see ./docs/layout.md.

• apps/NeuralFusionCore/lib/*.py — datasets, models, training loops, utilities, backtesting
• apps/ChronoBridge/scripts/data_ingest_service.py — fetch OHLCV/news and push to Redis
• apps/ChronoBridge/scripts/features_service.py — build features & fused embeddings
• apps/NeuralFusionCore/scripts/train_service.py — train model
• apps/NeuralFusionCore/scripts/finetune_service.py — incremental fine‑tuning
• apps/NeuralFusionCore/scripts/prediction_service.py — scheduled inference (writes to MongoDB/Redis)
• apps/NetWeaver/src/services/netweaver_train_service.py — train graph model
• apps/NetWeaver/src/services/netweaver_prediction_service.py — graph inference and ranking/graph outputs
• scripts/alphafusionnet_service.py — fusion + guardrailed reasoning + persistence
• scripts/metric_live_service.py — live monitoring metrics
• scripts/metric_monthly_service.py — periodic summary metrics
• scripts/metric_backtesting.py — historical validation runs (non‑advice evaluation)

---

Dependencies

• Python 3.11+
• PyTorch 2.x

---

Appendix

Upstream repositories

• NeuralFusionCore: https://github.com/Novoxpert/NeuralFusionCore
• ChronoBridge: https://github.com/Novoxpert/ChronoBridge
• NetWeaver: https://github.com/Novoxpert/NetWeaver

---

Authors & citation

Developed by the Novoxpert Research Team.
If you use this repository or build upon our work, please cite:

Novoxpert Research (2025). AlphaFusionNet: LLM‑Guided Neural + Graph Market Monitoring Engine.
GitHub: https://github.com/Novoxpert/AlphaFusionNet

---

Support

• Issues & Bugs: https://github.com/Novoxpert/AlphaFusionNet/issues
• Discussions: https://github.com/Novoxpert/AlphaFusionNet/discussions