Do you mean that it doesn't generate code that uses the shared libraries or that it's actually breaking in some fashion? Assuming you mean it's not taking into account your shared libraries, that's expected. The context that's currently provided in the prompt is quite limited, and the LLM really only has the provided context as insight into your project/org at the moment. There's already an enhancement request to be able to provide additional context with the requests: bitbucket.org/RoseSilverSoftware/illuminatedcloud/issues/2647 and in the future, Salesforce plans to make the LLM automatically aware of the respective org's contents without the API client having to include it as context in the request. If, on the other hand, you're seeing a bug, please file a bug in the public issue tracker.
@@vanbeukeringfamily1724 got it. Yeah, that's as-expected right now. I'll probably look into opportunities for context improvement a bit once I get the Winter '25 updates out, but the longer-term answer is that Salesforce is supposed to be adding full RAG support (effectively full knowledge by the model of the org's contents) at some point. Until then, the model only knows what it's been trained on and what you tell it in a given prompt.
Great education material! Thank you so much!
After this video I will be less afraid to show demos on AI models😉
Great video! I can finally try to use AI now.
Hi Scott thanks for the detailed Demo. Find that in standardised orgs with strong re-use via shared libraries it really doesn't work at all.
Do you mean that it doesn't generate code that uses the shared libraries or that it's actually breaking in some fashion? Assuming you mean it's not taking into account your shared libraries, that's expected. The context that's currently provided in the prompt is quite limited, and the LLM really only has the provided context as insight into your project/org at the moment. There's already an enhancement request to be able to provide additional context with the requests:
bitbucket.org/RoseSilverSoftware/illuminatedcloud/issues/2647
and in the future, Salesforce plans to make the LLM automatically aware of the respective org's contents without the API client having to include it as context in the request.
If, on the other hand, you're seeing a bug, please file a bug in the public issue tracker.
@@Illuminatedcloud no bug, but no org specific awareness of things built before. I
@@vanbeukeringfamily1724 got it. Yeah, that's as-expected right now. I'll probably look into opportunities for context improvement a bit once I get the Winter '25 updates out, but the longer-term answer is that Salesforce is supposed to be adding full RAG support (effectively full knowledge by the model of the org's contents) at some point. Until then, the model only knows what it's been trained on and what you tell it in a given prompt.