suchintan

10 ms·

91.
by suchintan 2y ago
I suspect accuracy is going to vary between domains that LLMs are being applied to Interesting call about validating this against YAML / TOML / CSV / HCL. HTML was natural for us because we're dealing with web data, so t
92.
Representing LLM inputs as HTML instead of JSON to cut input tokens by 11%(blog.skyvern.com)
15 points by suchintan 2y ago|2 comments
93.
by suchintan 2y ago
I've been using briefer for the past few weeks and have a few questions: 1. Do you intend Briefer to be a jupyter replacement or a metabase replacement? 2. Do you have any plans to open source briefer?
94.
by suchintan 2y ago
I love love love this idea! The papers all look like they're from April? Any chance to get more recent stuff?
95.
Analytics and attribution for open source projects(blog.skyvern.com)
7 points by suchintan 2y ago|0 comments
96.
by suchintan 2y ago
https://github.com/Skyvern-AI/skyvern This is pretty much what we're building at Skyvern. The only problem is that inference cost is still a little bit too high for scraping, but we expect that to change in the ne
97.
Transitioning from Engineers to Founders (+ the mistakes we made along the way)(blog.skyvern.com)
8 points by suchintan 2y ago|0 comments
98.
by suchintan 3y ago
I think there are some they're good enough at today. Auto generating meeting notes + AI context, auto responding / following up to emails, filling out forms (we do that pretty well at Skyvern with high accuracy)
99.
by suchintan 3y ago
I don't think ai agents are good enough to replace every job today, but they're starting to nip at the more junior / menial knowledge jobs I've seen a lot of success come from AI sales agents, just doing basic SDR style
100.
Aftermath of being on the front page of Hacker News for a day(blog.skyvern.com)
5 points by suchintan 3y ago|0 comments
101.
by suchintan 3y ago
We're definitely planning this. Skyvern currently intercepts all network calls that gets made -- we save them all as a HAR file for debugging purposes... but never looks at them A lot lot lot of scraping use-cases become simpler if you
102.
by suchintan 3y ago
We've discussed generating automation code internally a bunch, and what we decided on is to do action generation and memorization, instead of code generation and memorization. They're not that far apart conceptually, but there is
103.
by suchintan 3y ago
What do you mean? All of our code can be found here https://github.com/Skyvern-AI/Skyvern
104.
by suchintan 3y ago
Imagine that exact use-case -- pulling up relevant tax information and filling it Now imagine it from the accountant POV, where they have the same use-case for hundreds of clients This is where we've seen something like Skyvern really
105.
by suchintan 3y ago
This is a great point. This is something already on our roadmap. We call it "prompt caching", but I realize writing this that it's a terrible name. Will update! ( https://github.com/Skyvern-AI/Skyvern?tab=
106.
by suchintan 3y ago
Interesting. We've decoupled navigation and extraction for specifically this reason, but I suppose decoupling selector with input could let us use cheaper smaller LLMs to "select" and answer We've been approaching it a l
107.
by suchintan 3y ago
We're gonna build a self-serve UI soon! We just wanted to get it into people's hands ASAP :) Feel free to email me at suchintan@skyvern.com -- I can let you know when the self-serve UI is live
108.
by suchintan 3y ago
This can definitely be used for front end testing. Just tell it to do something like a user and monitor whether it's successful or not Here's a prompt example to try out { "url": " https://news.ycombinator
109.
by suchintan 3y ago
Actually this kind of stuff is super exciting -- we don't need to depend on companies exposing APIs for their website -- we can just use something like Skyvern instead!
110.
by suchintan 3y ago
When you say "getting the data in the client's infrastructure", do you mean self-hosting the robots? or something else?
111.
by suchintan 3y ago
We DEFINITELY will. I think we're planning on pushing that next week -- we've been super excited about it Just created this: https://github.com/Skyvern-AI/skyvern/issues/72
112.
by suchintan 3y ago
Nice! Thanks for sharing this. We tried approaches like VimGPT before but found the rate of hallucinations to be a bit too high to be used in production. The sweet spot definitely seems to be to combine the magic of Dom parsing AND vision W
113.
by suchintan 3y ago
We're quite different than LaVague. LaVague passes in the entire HTML DOM to the LLM to help it generate XPaths and valid Selenium code. ( https://github.com/lavague-ai/LaVague/blob/main/src/lava
114.
by suchintan 3y ago
Thanks! We definitely experimented with V only (that's the dream), but there's too much context missing: 1. What's behind a select option? You don't know until you click it, which means you need another iteration. This s
115.
by suchintan 3y ago
Saw the launch yesteday. Love all of the excitement in the space! LaVague is all about generating selenium code to interact with a specific page, and do it step-by-step Skyvern is all about taking a simple instruction and converting it to a
116.
by suchintan 3y ago
https://github.com/Skyvern-AI/Skyvern/?tab=readme-ov-file#fe...
117.
by suchintan 3y ago
I love all of these ideas!! 1. You can set a "max steps" limit when you run it locally https://github.com/Skyvern-AI/skyvern/blob/d0935755963b017ed... We also spit out the cost for each step within
118.
by suchintan 3y ago
We have a simple API we're building as a part of our cloud offering. It's in private beta today -- if you'd like to check it out please email me at suchintan@skyvern.com and I'd be happy to chat Thanks for the feedback r
119.
by suchintan 3y ago
Ah cool -- we weren't familiar with OpenAdapt. Will check it out. One big decision we made was to focus on browser automations (instead of computer automation like Adept or OpenAdapt). The reason for this was that we wanted to leverage
120.
by suchintan 3y ago
thank you!!

More ›