Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Change the Prompt Message to use the "system" role for better accuracy.
I've found (at least with my experiments with PF2e) that if you send nothing for the system message and 2 user messages, you'll get less accurate responses where the system tends to invent rules that don't exist. This change should tighten up rules accuracy a little in PF2e, and hopefully shouldn't negatively impact other systems (a couple cursory tests with D&D 5e questions seemed just fine).
Here's the sample results I have for the 2 different approaches:
Using "user" for both messages, with the question: "What are the rules for grappling?"
Response:
Using "system" for the system prompt and "user" for the user's question:
Response:
The key important differences here are the addition of the immobilized condition, the explanation of that condition, and occasionally when using "user" ChatGPT will claim you can do things like "move" or "pin" a creature, which are not allowed actions in PF2e. I have yet to see said issue when using the "system" approach.