Ooh, yeah it's looking fairly effective. I asked some questions and I like that the answers address differences in language implementations (e.g. python bindings vs cpp) and that there are relatively good code suggestions. I assume this means the tests and benchmarks are helping the RAG a lot, so general well done to all contributors. Out of curiosity, are the cookbooks part of the learning/retrieval sources? Sent from Proton Mail for iOS On Fri, Mar 28, 2025 at 13:58, Neal Richardson <neal.p.richard...@gmail.com> wrote: Awesome, thanks for doing this! I tried out a few queries and was impressed
with what I got back. Neal On Fri, Mar 28, 2025 at 2:01 PM Nic Crane <thisis...@gmail.com> wrote: > Hi folks, > > The kapa.ai bot is now live on the dev docs! We're using it to test > whether > we can enhance user experience of our documentation via retrieval augmented > generation which generates answers to user questions by combining > similarity searches of the docs and LLM output. > > You can use it by clicking the "Ask AI" button in the bottom right corner > of https://arrow.apache.org/docs/dev/index.html > > Currently we've just ingested the docs, but as we experiment more, we can > add other sources like Zulip and the mailing lists if we think these will > provide useful additional source material. > > Please give it a go and use the "good answer" and "bad answer" buttons so > we can assess how useful this is likely to be for our end users; bonus > points for adding a better answer when you're not happy with the one given. > > I'll be checking in on it every so often and will report back here with the > results. > > Thanks, > > Nic >
signature.asc
Description: OpenPGP digital signature
