Knowledge Portal

Documentation Hub & IDS Semantic Explorer

Instant client-side semantic search across over 1,600 indexed documentation files, architectural contracts, PyTorch implementation guides, and hardware telemetry records.

🔍

Developer Quickstart

5-Minute Rapid Setup Guide

Get the entire ImpressionCore Model Builder and Runtime running locally on Windows, Linux, or macOS in 5 simple commands.

1. Clone & Setup Virtual Environment

Requires Python 3.10+ and CUDA 12.1+ (optional for GPU acceleration):

git clone https://github.com/kirklasalle/impressioncore.git
cd impressioncore
python -m venv .venv
.venv\Scripts\activate      # Windows
# source .venv/bin/activate # Linux / macOS
pip install -r requirements.txt

2. Launch System A & System B

Launch the Model Builder on Port 5000 and Runtime on Port 8000:

# Launch Unified Model Builder (Port 5000)
launch_builder.bat

# Launch ImpressionCore Runtime (Port 8000 / 5173)
launch_impressioncore.bat

ImpressionCore Live System Requirements Checker

Automated System Requirements & Compatibility Checker HARDWARE AUDIT

Quality Assurance

Automated Verification & CI/CD Testing

ImpressionCore includes a comprehensive automated builder verification script testing all 9 builder site functions, preflights, and real model runs.

Terminal / Command Prompt Comprehensive Test Suite

# Run the full automated Builder verification suite:
python src/dev_tools/exercise_builder_site.py

# Run PyTest unit tests across core modules:
pytest tests/ -v

✅ Preflight Checks

Automated CUDA driver checks, device memory queries, and PyTorch 2.6+ bypass validation.

✅ Model Contract Audits

Forward pass verification for B1 39M, B2 50M, and B3 504M architectures with zero NaN gradients.

✅ Non-Repudiation Logs

Automated SHA-256 checkpoint hashing ensuring exact mathematical reproducibility.