• Train a deep transformer on your data — then chat with it.
• No cloud. No API key. Pure Python + optional matplotlib.

---

Session Flow

  1. Choose input format  [1]TXT  [2]CSV  [3]Excel  [4]JSON  [5]TSV  [6]Auto
  2. Enter file path
  ↓  Training starts — 5 LIVE windows:
       W1 Loss Pulse   W2 Neural Flow (animated!)   W3 Token Heatmap
       W4 Gradient Health              W5 Live Output
  ↓  Chat window opens automatically
  ↓  Ask questions from your data  |  Out-of-scope queries are refused

Quick Start

git clone https://github.com/Chintanpatel24/Neuryx_v2.git
cd Neuryx_v2
python3 neuryx_v2.py

Optional but recommended — for charts

pip install matplotlib openpyxl     # optional — for graphs
sudo apt-get install python3-tk     # optional — for chat GUI (Linux)
python neuryx_v2.py

Sample Datasets (included)

File	Format	Use
data/sample_names.txt	TXT	Name generation
data/sample_weather.csv	CSV	Weather Q&A
data/sample_logs.json	JSON	Log event queries
data/sample_events.tsv	TSV	UI action queries
data/sample_text.txt	TXT	Phrase Q&A
data/sample_sequences.xlsx	XLSX	Sequence data

CLI Flags

python neuryx.py --train FILE --steps 600 --mode char
                 --depth 64 --rifts 4 --horizon 128
                 --temperature 0.35 --no-live --no-chart

Architecture

• core/flux.py — scalar autograd engine (no ML deps)
• core/lattice.py — 4-block causal transformer (depth=64, streams=8)
• core/retriever.py — TF-IDF retrieval + out-of-scope detection
• render/livewire.py — 5 separate live matplotlib windows
• render/chatwin.py — Tkinter chat window (terminal fallback)

Requirements

Python >= 3.10
matplotlib  (pip install matplotlib)     # graphs
openpyxl    (pip install openpyxl)       # Excel support
tkinter     (apt-get install python3-tk) # chat GUI

MIT License.