[syx]

For those wondering about the use case, this is very useful when enabling streaming for structured output in LLM responses, such as JSON responses. For my local Raspberry Pi agent I needed something performant, I've been using streaming-json-js [1], but development appears to have been a bit dormant over the past year. I'll definitely take a look at your jsonriver and see how it compares!

[1] https://github.com/karminski/streaming-json-js

[rokkamokka]

For LLMs I recommend just doing NDJSON, that is, newline delimited json. It's much simpler to implement

[rictic]

Do any LLMs support constrained generation of newline delimited json? Or have you found that they're generally reliable enough that you don't need to do constrained sampling?

[stevage]

I love NDJSON in general. I use it a lot for spatial data processing (GDAL calls it GeoJsonSeq).

[cjonas]

Particularly for REACT style agents that use a "final" tool call to end the run.