Great read as always! Point for thought: would PyArrow (or just broadly Apache Arrow) not provide sufficient, fast handling of raw Parquet to not then require a secondary format (Lance)?
Thanks for the insight. As an enterprise architect, I am always looking for ways to hedge technology investment against functionality. This article, which I need to chew on for a bit, is interesting. Thank you.
Hi, great read, indeed, as always. Could you source your token marketshare by model author bar chart ? I googled for it and found no results
regards
Hi Laurent, thanks. You are right I should have included it: https://x.com/OpenRouterAI/status/1923429107234202101
Great read as always! Point for thought: would PyArrow (or just broadly Apache Arrow) not provide sufficient, fast handling of raw Parquet to not then require a secondary format (Lance)?
Yes arrow for in memory processing but you need to query data from an efficient storage format. That's where Lance adds value.
Thanks for the insight. As an enterprise architect, I am always looking for ways to hedge technology investment against functionality. This article, which I need to chew on for a bit, is interesting. Thank you.
Thanks Ed !