NeuroMANCER GPT

This folder contains a script that is intended to format the contents of the NeuroMANCER repository in a way that is suitable for ingestion in RAG-based "LLM-assistant" pipelines.

When run, this script will produce files that can be ingested/indexed by RAG-based pipelines representing specialized contextual knowledge to inform questions about NeuroMANCER.

We intend this script and its outputs to be used by the community to leverage LLM-assistants where desired. We hope that by providing this script, we lower the barrier to entry for using LLMs to assist the use, development, and understanding of NeuroMANCER.

Dependencies

The following pip packages are required by the script:

• pypandoc
• panflute
• tqdm

They may be installed using:

pip install pypandoc panflute tqdm

Usage

When run, i.e., by executing

python3 ingest.py 

the script generates the following files:

• knowledge/docs.txt concatenated NeuroMANCER documentation.
• knowledge/examples.txt concatenated NeuroMANCER examples, as python code.
• knowledge/src.txt concatenated NeuroMANCER source code.

These files represent the contents of the NeuroMANCER repository, formatted in a way that an LLM should be able to parse them more easily. An expected use-case is for these files to be used in a RAG-based pipeline, which as of writing have been developed to the point that using them should be as simple as moving/uploading them the appropriate directory/service of a local. These files should be useable with both open-source software for RAG that can be run locally and with commercial "RAG-as-a-service" products.

Together with these files, we provide a base "system prompt" for such a pipeline:

• knowledge/prompt.txt The system prompt for an AI assistant