johngunderwood/atlas-search-tuner

Query Tuner for Atlas Search

atlas-search-tuner's Issues

Separate schema from index definition retrieval

Currently schema and index definition are retrieved together when the user connects. Only the schema is required for the 'Index Builder' tab while the index definition needs to be retrieve for the 'Query Tuner' tab. For the index definition the app should be able to retrieve based on an index name provided by the user separately from initial connection check.

Improve connection useability

Two options:

1. Use drop down for database and collection specification (many changes)
2. Change database and collection input so it is just a namespace string: "." (few changes)

Improve the preview of 'search fields'

“what shows up in the results” is mostly not about field mappings, it’s more about “which fields am I projecting”.

If I think about setting up my UI to search for movies - I’d want a movie poster to show up in results, but I’m probably not searching in the movie_poster field…. The search box should be about field mappings, i.e. which text fields I’m searching in.

This could possibly be solved by using highlighting to display hits. It needs to be clear to the user that they are not building the UI experience, but the search behaviour

Add a result re-ranking feature

Provide a way for a user to re-rank results and persist the ranking rules back to Atlas.

Fix tricky dropdown on field selection

In the index builder modal the dropdown selector is trickt to dismiss.

Possibly an issue with leafygreen

Make it possible to have multiple filters

Fix bug with failed connection error message when password has '@'

When a connection fails the error message can be misleading due to naively splitting the url on '@' character

SearchResultsFields component cannot handle array of documents

Once index builds the UI doesn't change to add Query Tuner

Once index build is complete, there's no Query Tuner. And there's still a Deploy button.

Add $searchMeta facet query to tuner

Provide a way for users to generate $searchMeta queries to return facets

Handle nested fields when creating search index

Currently there is no logic to correctly create nested fields using the 'document' type. This needs to be added.

Facet querying is broken. searchMeta stage is not built

Update README to reflect new app design

Query tuner fails if no facets are defined

"TypeError: Cannot read properties of undefined (reading '$searchMeta')"

Make duplicate index error clearer | Make it clearer that the user can't create a duplicate index name

Thoroughly test user permutations to make sure state changes are working

Improve user journey around index creation

When a new index is building the user should be able to close the modal and come back to it.

OR the user should be prevented from closing the modal.

Currently the callback flow breaks if the modal is closed.

Provide 'build new index' option now that index build variables are in main scope

After initial index build the user needs a way to start fresh

Increasing the limit for categories

I wonder what algorithm we're using for categories/facets. For this test dataset I don't have brands showing up and I still want to do facets on them. But there are more than a 1000 brands here...

To get the data.

git clone https://github.com/Ksushis/grocerystore

mognorestore from terminal

cd grocerystore
mongorestore

You should now have grocery database with items collection with 19564 documents

Add configure button to search bar

search bar doesn't have configure button, you have to click on it to summon the modal where you specify autocomplete

Would be good to have it have it's own configure button like other elements

Improve 'connect to index' UX

It's confusing to the user - what does it mean to 'connect'? We now have two 'connect' buttons in view.

Selecting an index sets the variable. This can happen in the background and we don't need a connect button.