Orchard Core and MCP

[agriffard]

Goal: How a GPT, Copilot, "you name it" LLM tool, ... can have better answers on questions about Orchard Core?
The answer wil always be: It needs more context.

What are your approaches to achieve this?
I guess some of you will come up with MCP (Model Context Protocol).

Has anyone made some further researches and would be able to make a demo during a meeting?

Anyone knows about https://context7.com/orchardcms/orchardcore ?

#MCP all the things.
#MCP is the new AI magic wand.

[MichaelPetrinolis]

I have defined a knowledge base in Copilot with our private repos and public OrchardCore repos. This way a dev can use @github #kb knowledgebase when asking questions. The context is enriched with all the info from the markdown files from the sources(docs). If you also configure the copilotinstructions.md you add additional context. Unfortunately they have not yet integrated the knowledge base in the copilotinstructions.md, so the @github agent is required

[Piedone]

Are you referring to this feature? https://docs.github.com/en/enterprise-cloud@latest/copilot/how-tos/context/creating-and-managing-copilot-knowledge-bases Sadly it needs an Enterprise plan.

[MichaelPetrinolis]

You are right, I missed that. Those who own enterprise subscription will benefit from the knowledge base feature. The discussion is about using Copilot to implement a site or use AI features from within Orchard Core dashboard?
Although I have not used Mike's AI modules, it seems that fit perfect for adding custom tools for use within the dashboard.
Developing theme/custom modules from IDE, needs context on how to create vital OC elements. Drivers, Views (liquid and razor), use tags, liquid extensions etc. A copilot-instructions.md in the solution, would guide the community what additional context they should add.

[Piedone]

Mike's AI modules include an MCP client, but I guess similarly a server that exposes data from an OC site could be implemented too.