how can i attach a domain specific llm?

[curlyfro]

very cool but suppose my work is domain specific. how can i attach a domain specific llm? is there a way for me to attach semantic kernel or langchain for .net? this would give me choice of my backend

[SteveSandersonMS]

Yes, you can customize the inference in various ways:

• You can use any other inference REST endpoint if it's compatible with OpenAI (e.g., Ollama): https://github.com/dotnet-smartcomponents/smartcomponents/blob/main/docs/configure-openai-backend.md#using-ollama-or-another-self-hosted-endpoint
• You can implement IInferenceBackend in your .NET code and put in any logic you want to call your domain-specific LLM
• You can subclass component-specific inference services and put in any logic you want to call your domain-specific LLM, e.g. https://github.com/dotnet-smartcomponents/smartcomponents/blob/main/docs/smart-paste.md#customizing-the-prompt