This is what attracted me to CustomGPT and away from ChatPDF.
@alden seems to have a solid vision of a fully automated platform. I cannot speak for the creator, but I suspect in the near future …
- It will be possible to arbitrarily add or remove documents from the training set through the CustomGPT API.
- It will be possible to replace documents in the training set through the API.
- It will be possible to enhance the training set with JSON and JSONL documents as needed to improve inference performance.
Already doing this with CustomGPT. I use Google Apps Script to combine content from many sources to dynamically build text and PDF documents that serve as training sets. Ideally, CustomGPT’s projects will allow the dynamic inclusion of other training documents the way it presently supports site maps (over HTTP). I think the creator is totally in alignment with agile pathways to support content worthy of training inclusivity.
