Rossi

> I think it would be necessary to implement something to choose the best text and mark the others as versions Yes, sure. We usually choose the best formatted content...

Each of those would be a OpinionContent row If we have an Opinion with Opinion.xml_harvard, Opinion.html_columbia and Opinion.html populated, that single Opinion would produce 3 OpinionContent rows; each row would...

About adding a new "Scanning" source to OpinionCluster.source and Docket.source; I think we definetly should About this > what do you think about having a constraint to prevent OpinionContent objects...

@quevon24 > OpinionContent will not be used at the moment to show information to the user, so it is necessary to make the opinion field optional in the model I...

> That'd catch any kind of error we have, assuming that the long ones don't do a partial transcript or something. @mlissner I think this can be handled by using...

I downloaded the details for 1236 events / logs from Sentry, which map to 486 unique dockets. Then, I manually inspected each court group, and found that some are indeed...

Other courts that need special lookups for docket matching: - `ny` Example [docket](https://www.courtlistener.com/api/rest/v3/dockets/1647630/), with docket number "No. 86", has 2 clusters assigned, one from [1983](https://www.courtlistener.com/opinion/2588854/s/), another from [2024](https://www.courtlistener.com/opinion/10118643/s/)

I finished breaking this issue up (for the most apparent courts, there may be some stragglers) - the title is no longer correct (it was for ariz, fladistctapp and ohioctapp...

Besides further improvement, this one will have to wait until we [finish](https://github.com/freelawproject/courtlistener/issues/6363) populating the column

@albertisfu I think this is ready for review