feat: use latest create llama and llamaindex

run-llama / chat-llamaindex

https://chat.llamaindex.ai

MIT License

869 stars 265 forks source link

feat: use latest create llama and llamaindex #101

Closed thucpn closed 3 months ago

thucpn commented 3 months ago

New features:

upgrade to latest create-llama and fix import
upgrade to latest LITS and fix issues related to tiktoken
update /upload API to receive datasource in payload
update UI to pass datasource to payload of /upload API
able to see all datasources in bot config section
add filter for ContextChatEngine

Change in flow:

when a new bot is created, its datasource name is bot_${botId}
when datasource is not found, getDataSource function will return a new index with empty nodes

Note: use latest create-llama on PR: https://github.com/run-llama/create-llama/pull/202

vercel[bot] commented 3 months ago

@thucpn is attempting to deploy a commit to the LlamaIndex Team on Vercel.

A member of the Team first needs to authorize it.

vercel[bot] commented 3 months ago

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated (UTC)
chat-llama	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	Aug 5, 2024 1:33pm

marcusschiesser commented 3 months ago

@thucpn there are two blockers from merging this:

If I just send "Hi" to "My documents" I get the following error:

The reason is that we need a datasource for "My documents" (we always should have one defined) - so add a "private" datasource
File upload is leading to a 500 error. This is expected as the datasources are read/only and the persist call will obviously fail. For this PR, we just need to show an error message

The benefit of this PR will then be: People can locally run chat-llamaindex and use the file upload feature (as in dev environments the indexes are writeable)