I have an email inbox with about 100 emails I’m looking to “chat” on. How could I go about loading the emails onto vectara in either a plaintext format or a .eml format? I would like to use langchain if needed and hook up the Vectara search to a user friendly GUI.
Im also wondering if it is extensible / scalable… say for example I wanted to eventually incorporate web browsing capability via an agent.
TLDR : Trying to load a specific email inbox contents onto vectara preferably in a dynamic way as emails continually come in, and then search that inbox in a user friendly GUI.
It’s pretty straight forward to load a plaintext email into Vectara. I would recommend that you just use the Standard indexing API to index those emails, each one can be its own “document” in Vectara and the Vectara system will break it down into proper chunks automatically.
If you want to incrementally add new emails as they come, you can keep track of the emails you uploaded and just “index” the new ones as they appear.
If you want to use LangChain to upload the email content, please see here for an example / instructions for how to load data via the LangChain integration.
For UI:
You can build a question-answering user interface relatively quickly using vectara-answer
here is an example to build a chat application using LangChain with Vectara.
Note that these are in the Python version of Langchain, and you can of course use the LangChainJS version if you prefer Javascript, and Vectara is also integrated with Flowise if you prefer a drag-and-drop interface. Let me know if I can help further.