Knowledge Engine for Houdini (KEH)

Status: Setup Phase

Warning

Work in Progress: This project is currently under active development. Some modules, links and files may be placeholders.

A specialized RAG-based system designed to index, structure, and retrieve SideFX Houdini documentation.

🏗 Architecture (Src Layout)

src/
    keh/                # Main Package
        core/           # Analytical (Brain)
        ingestion/      # Input (Parsers)
        factory/        # Transformation (Factories)
        interface/      # Communication (UI)
        service/        # Service (Telemetry)
        storage/        # Data Layer
docs/
    adr/                # Architectural Decision Records
    knowledge_base/     # Operational Guides
    corpus/                 # (GitIgnored) Raw & Processed Data

🧠 System Overview (The "Truth Engine")

KEH is not a simple chatbot. It is a Reference-Checked RAG System composed of 5 isolated contours:

• Ingestion: Parsers normalize Houdini help files.
• Factory: Converts data into Vectors and Graph nodes.
• Brain: Uses a Zero-Hallucination Feedback Loop to audit every citation.
• UI: Embeds directly into Houdini via PySide6.
• Telemetry: Monitors VRAM to prevent system crashes.

Read the Core Architecture Specification

Documentation and Licensing Notice

• KEH is an experimental personal research tool for improving navigation of locally available Houdini documentation.
• It is not affiliated with, endorsed by, or sponsored by SideFX.
• The project does not aim to redistribute, republish, or replace official SideFX documentation. Its intended use is citation-based retrieval that points users back to official documentation sources.
• Any documentation ingestion should respect SideFX licensing terms, copyright, and access restrictions. If SideFX provides official guidance, API access, or restrictions for documentation indexing, this project should follow them.

🛠 Technology Stack (See ADR 004)

• Language: Python 3.11 (Managed via keh-env)
• Databases: Qdrant (Vector), Neo4j (Graph)
• Orchestration: Celery (Queue), Redis (Broker)
• User Interface: PySide6 (Embedded in Houdini)
• Infrastructure: Docker Compose
• Quality Assurance: pip-tools, black, flake8, bandit, pytest

Read the full Technology Stack Decision (ADR 004)

🚀 Getting Started

1. Environment:

   conda activate keh-env
   pip install pip-tools
   pip-sync

2. Configuration:

   • Launch kehHoudiniPanel (See Interface Documentation).
   • Set Houdini Install Path (e.g., C:\Program Files\Side Effects Software\Houdini 21.0.596).
   • Click "Initialize Corpus" to trigger automated hydration.

📜 Protocols

• Development: All code must reside in src/keh.
• Commit Messages: British English, Imperative Mood.
• Safety: "Zero-Hallucination" citations required for all outputs.

Changelog

Week of 2026-02-02 (Foundation & Standards)

• Project Launch: Effective Start Date: 2026-02-06
• Jira:
  • Initialized project workflow.
  • Completed Project Initialization phase (Task KEH-1); 4h 30m logged.
• Architecture:
  • Defined foundational stack (Python, Qdrant, Docker) and rigorous verification strategy.
  • Deployed Source Layout and architectural documentation structure.
• Environment & DevOps:
  • Configured isolated development environment (keh-env).
  • Established Readme-driven documentation standard.
• QA & Security:
  • implemented strict dependency locking via pip-tools.
  • Integrated security scanning (pip-audit) and pre-commit hooks.
• Documentation:
  • Consolidated project context and architectural vision.
  • Refined internal operational workflows and directory scanning protocols.