List of all documented functions for AI consuption

[Eric Lendvai]

I exported all the Harbour functions documentation from Harbour.wiki in a format that could help AI when you are requesting to generate Harbour source code.

ChatGPT for example would make up some Harbour functions that did not exists.

The following repo:  https://github.com/EricLendvai/harbour-language-for-ai-training

has a yaml file listing all the functions with categories.

I also have this zipped up so it could be sent to ChatGTP directly in a chat session.

Currently ChatGTP, most of the time, cannot read github repos when you are in the middle of a chat.

Will see if this will help, either by sending the file, or for future AI Model training.

[berg...@gmail.com]

Thank you so much. Otto

[roberto....@gmail.com]

some time ago I uploaded to Gemini the (partial) documentation of my own classes and libraries and an example of how I implement a typical 'view object' of a table (browse,edit,search,report classes). Then I asked to add a button to the standard buttonbar to call a custom function working on the table. It was a simple function and Gemini built it 'almost' correctly. I think it would be posible to create a custom gpt to help writing functions and classes using your own libraries and functions and holding your style at the same time. What do you think about?

[Eric Lendvai]

Thanks Roberto,

Following your comment I asked ChatGPT " If I use your GPTs feature, could this YAML of Harbour functions be remembered there ? "

And luckily it can. Here is its full answer:

Here’s the reality for custom GPTs (the “GPTs” feature inside ChatGPT):

🔹 What a GPT can remember

• A custom GPT can have its own “custom instructions” (like a built-in memory note).

• You can embed persistent textual knowledge there — e.g.,

  “This GPT uses the official Harbour function YAML dataset from https://github.com/EricLendvai/harbour-language-for-ai-training to ensure all code generation references only known functions.”

• You can even upload small files (< 20 MB total) as part of the GPT’s knowledge base.
  Those files remain stored with that GPT and are reloaded automatically for every session.

🔹 What it cannot do (yet)

• It cannot dynamically sync with your GitHub repo unless you re-upload or re-link the ZIP yourself.

• It cannot run the validator code or re-parse YAML automatically at runtime (it can only reference pre-uploaded knowledge).

So:
✅ Yes, you could embed your YAML file (or a compacted JSON form of it) in your GPT’s knowledge base — that would make it persist across all sessions and enforce “no invented functions.”
🚫 But it won’t auto-update if you push a new revision to GitHub; you’d need to re-upload the new version to that GPT.