feature: enhanced docs search

pingcap / tidb.ai

https://TiDB.AI is a Graph RAG based and conversational knowledge base tool built with TiDB Serverless Vector Storage and LlamaIndex. Open source and free to use.

https://tidb.ai

Apache License 2.0

193 stars 25 forks source link

feature: enhanced docs search #320

Closed sykp241095 closed 2 weeks ago

sykp241095 commented 2 weeks ago

Currently we only support search by URL, it doesn't make sense: https://tidb.ai/documents

We should enable searching with name, tille, url or any reasonable fields. and also with enough filters like: mime_type

Icemap commented 2 weeks ago

Hi @sykp241095. Would you mind describing this feature more specifically? Now, we can only search the URL for sure. And in the database, we have some other fields, which are:

created_at
updated_at
name
content
mime_type
last_modified_at
meta
index_status

Which fields should we add to the enhanced doc search API?

sykp241095 commented 2 weeks ago

@Icemap we can do it with best practice? you can ignore my suggestions.

(I am not sure if mimetype is a ENUM or not)

Icemap commented 2 weeks ago

Yes. mime_type is an enum type. Aside from that, I added created_at_start, created_at_end, updated_at_start, updated_at_end, last_modified_at_start, last_modified_at_end, name, mime_type, index_status, language, and product filters. Also, change the blur name query to source_uri. @sykp241095 PTAL.

Icemap commented 2 weeks ago

Feature completed. If you have any questions, please feel free to reopen this issue.