3. Standard vs. RAG Version
This demo is available in two flavours.
The "standard" demo uses OpenAI's ChatGPT (coupled with an on-cluster Weaviate cache) to look up destination advice for any destination.
The "RAG" version (available on the rag branch) will only produce destination advice for places the system has explicitly been trained on (the files in the destinations folder on the rag
branch). Namely, Bali
and Sydney
.
This is achieved by:
- Reading each file from disk when the app starts
- Sending the contents of the bali and sydney HTML pages along with each request and explicitly telling the model to only use the information provided in those documents.
The RAG version of the demo mimicks training an LLM on an internal knowledgebase.