jasonwcfan

20 ms·

61.
by jasonwcfan 4y ago
You can use any embeddings you want! We normally use OpenAI's ada which is $4 per 10 million tokens, which is fine for now. But eventually we'll need to figure out a way to incrementally sync data from SaaS tools instead of re-vec
62.
Show HN: Open-source ETL framework to sync data from SaaS tools to vector stores(github.com)
63 points by jasonwcfan 4y ago|8 comments
63.
by jasonwcfan 4y ago
Congrats on the progress! Generative AI is all the hype now and there's a few startups exploring using AI to generate front-end code. Is this on your roadmap as well?
64.
by jasonwcfan 4y ago
1) We could, yes! It just hasn't been a priority yet 2) If the docs are open source you just need to provide the url to the github repo where they're stored and we will be able to ingest them. Submitting PRs to update the docs wil
65.
by jasonwcfan 4y ago
That's clever! So it looks like rather than use a vector store, in your approach you use GPT and the table of contents to find the right section? How do you find it performs in terms of accuracy and cost?
66.
by jasonwcfan 4y ago
I see - we'll reconsider then! Appreciate the feedback
67.
by jasonwcfan 4y ago
We used Screen Studio! https://www.screen.studio/ Great product
68.
by jasonwcfan 4y ago
Awesome! Great to hear :)
69.
by jasonwcfan 4y ago
Yep! It could be, though we had to be be very focused on the channel so we started with Slack and Discord. Personally I think there's less value in an interface over the docs vs a community. Engineers might prefer faster answers from d
70.
by jasonwcfan 4y ago
Good point - I suppose we'll cross that bridge when we get there.
71.
by jasonwcfan 4y ago
Being a platform, this is just one of a thousand different things they could prioritize. Sure they could cut us out and release a feature that searches through Github docs, but frankly there are much more valuable use cases of generative AI
72.
by jasonwcfan 4y ago
Never bullshit, but like any application built with GPT it does occasionally hallucinate. We published info on how we solved that problem here: https://medium.com/p/f3bfcc10e4ec We've also made some improvements i
73.
by jasonwcfan 4y ago
Thanks for the feedback! It was just something we added for our Product Hunt launch since that community loves promo codes. You can use the code SIDEKICKPH2023 for the 30 day free trial :). The code works until the end of the week!
74.
by jasonwcfan 4y ago
We might if we hear a lot of people asking about it! Would you use it if it supports Teams?
75.
by jasonwcfan 4y ago
Thanks for the support! :)
76.
by jasonwcfan 4y ago
If we force content_contains_answer to false, we always re-run the evaluation step first to get a new justification for why the content doesn't have enough information. Then we run that justification through another prompt to convert
77.
by jasonwcfan 4y ago
We actually migrated from Pinecone to Weaviate and it's been great far and I definitely recommend it. Top 2 features are: 1) the way they handle class properties is super intuitive and allows you to configure which properties to index,
78.
by jasonwcfan 4y ago
We started with external docs because it was easier to ingest them add make it useful immediately, but we're definitely thinking about internal use cases too. In theory this should work just as well on Confluence as on Github - we&#x27
79.
by jasonwcfan 4y ago
LOL I agree that's a pretty good name!
80.
by jasonwcfan 4y ago
Yeah - we think similarly to how Gong and other sales enablement products helped companies quantitatively measure and improve sales rep performance, there will be a lot of value in products that help measure and improve customer service per
81.
by jasonwcfan 4y ago
Exactly. I didn't mention it in the post but the master plan is to give companies analytics over their customer conversations and documentation. This was technically possibly years ago but no one did it because customer support was see
82.
by jasonwcfan 4y ago
We have! The reason we didn't start there is because Confluence is usually used for internal docs, which means a longer sales and adoption cycle. The nice thing about developer docs is that they're public so we can ingest them and
83.
by jasonwcfan 4y ago
Yeah we're thinking any type of product in a technically complex field could benefit. Legal, compliance, software, etc. For these types of products companies need highly trained employees with specific expertise to provide support and
84.
by jasonwcfan 4y ago
We started out with all docs actually, and decided to narrow to dev docs because 1. Markdown files are structured and easier to parse 2. Engineering teams tend to have more influence in their orgs. We tried talking to a lot of traditional c
85.
by jasonwcfan 4y ago
Thank you! Appreciate the support
86.
by jasonwcfan 4y ago
It was great feedback. We've been deep in the weeds building so it's hard to pull ourselves up out of the weeds, but for something that reads from and updates documentation it's probably easier to see it in action than than t
87.
by jasonwcfan 4y ago
We're also considering adding the bot to Gitter and Discourse but frankly just haven't had the time to build out those integrations yet! It just happened that some of the first companies we reached out to used Discord. One of them
88.
by jasonwcfan 4y ago
Thanks for the suggestion! Just updated: https://www.getsidekick.ai/
89.
by jasonwcfan 4y ago
Hey thanks for asking! 1. If someone reacts to a message in Slack with the :memo: emoji, Sidekick will take the contents of the message and search for a chunk in the docs that closely matches the content. Each chunk is tagged with the sourc
90.
Launch HN: Sidekick (YC W23) – AI support bot for developers
158 points by jasonwcfan 4y ago|87 comments

More ›