Open mpr1255 opened 2 months ago
Hey, it is unlikely for lnx to support JSON fields, in fact in future versions it is likely that dynamic schemas will be explicitly disallowed with a focus on strict schemas with the ability to update the schema on the fly at your convenience.
Ah. Thanks for such a speedy response and being clear on the plan. Yes, my use case is a bit cruddish -- search backend for an LLM app. So users will do transformations on a field and save it back to the db in another field that then needs to be searchable. So the second scenario you mention would work. However I guess neither is implemented now anyway. We should probably look at ParadeDB or quickwit then I suppose. Too bad because I loved the yt vids. Such a good vibe from this software.
For that sort of application if you are mutating and changing document data to some unknown/arbitrary set of fields you may be better off using Meilisearch if you need typo-tolerance or strict low-latency. Otherwise yes Quickwit probably would be a good choice :slightly_smiling_face:
Thanks SO much for the rec to Meilisearch. It had been vaguely on my radar but I hadn't done a deep dive. It is almost exactly what we're looking for in terms of features and data model. Fantastic. Most appreciated.
Tantivy enables the json type as an indexable field. It seems LNX does not. Anyone got any appetite to do that? Would unlock a huge number of use cases.