This repository contains a series of notebooks that simulate the corpus creation, data prep and ingest required to build a RAG powered chatbot.
To use this:
- Setup vectara-skunk-client YAML file: https://github.com/davidglevy/vectara-skunk-client
- Setup Jupyter Lab: https://jupyter.org/
The notebooks are in the notebooks folder.
- Corpus Creation: Done, embedded in notebooks
- Filter Attributes: Done, see 02 Presentation Corpus
- Data Loading: Done, each notebook loads it's data.
- Crawling: Outstanding; currently data is static in repo
- Webhook: Partial: working with Chatbot.com parameters/flow
- Upload to Corpus: Partial; I have webhook downloading Chatbot.com CDN but need to then push to Corpus
- Custom Prompts: Not Done, once we have actions working well, I'll introduce this.
- Memory: Not Done (beyond Chat API memory), I will add a memory of conversation to inject into custom prompt.
- Action Aspect: Not Done, I will add this soon to intercept conversation which expects an action.