SéanceNet: Offline Digital Séance for Forgotten Architectures & Languages

Description and use

SéanceNet lets developers translate modern code into legacy syntax locally, preserving computing history and ensuring privacy.

Problem

Modern developers rarely remember foundational systems; need a way to translate modern code into legacy architectures.

Outcome

SéanceNet allows developers to transform modern code into legacy syntax locally, preserving computing history and ensuring privacy.

Workflow

• 1. User selects a legacy spirit (CM-5, COBOL, FORTRAN-77).
• 2. User inputs modern code and optional incantation.
• 3. The system loads the Qwen2.5-Coder-1.5B-Instruct model locally.
• 4. The system constructs a prompt with the spirit persona and user code.
• 5. The model generates legacy code translation.
• 6. The output is displayed in a Gradio UI.
• 7. Transmutations are stored in MongoDB Atlas Vector Search for future retrieval.

Evidence

SéanceNet is an offline-first digital séance that channels the spirits of long-dead programming languages and vintage supercomputer architectures.

Source

The system loads the Qwen2.5-Coder-1.5B-Instruct model locally across GPUs.

Source

Persistent Memory with MongoDB Atlas Vector Search stores code transmutations in an Atlas collection.

Source