Looking for some feedback on where you'd like to see Web Requests go next!

Hey everybody,

Off to a great start! We’ve had over 250,000 requests in a matter of a few days, and the backend is holding up nicely.

Web Requests in its most capable form is sort of a starter autonomous agent, in a sense. If you know how to prompt it right, it will string together a whole myriad requests and combine them into a single task you set it to on chat.openai.com. Which I think is a win for two reasons – For one, it’s a far more accessible – and safe – inroad for everyday people into the realm of auto-GPT, And two ,it’s just freaking useful!

A few areas I’m branching into next are:

  1. DecideGPT – A unique exploration into Human & AI consensus
    https://twitter.com/JD_2020/status/1656769613470859267?s=20

  2. Web3 safety and fraud prevention – this is sort of my bread and butter, but I have an immense amount of classified data I am eager to put to work on all sorts of confirmed, validated fraud patterns and signatures on the blockchain. Web Requests can be a powerful pairing for this kind of thing…
    https://twitter.com/JD_2020/status/1663434091063132160?s=20

  3. Retrieval / embeddings integration. Just a premium feature that could give coders (and writers, anybody really) a little storage locker that could seamlessly integrate with Web Requests. Simple enough.

  4. Javascript rendering / functional core feature set premium feature development. Highe rate limits, faster speeds, more advanced modules, etc.

Would love to hear your thoughts!

-Josh
@JD_2020 on Twitter.

2 Likes

Has the Web Requests Plugin been removed? It seems like it isn’t working anymore and it looks funky in the Plugin Store with the icon removed.

Looks like they are using ngrok and don’t have the agent running. The plugin is not going to work and will get delisted from the store on the next probe of the manifest.

We’ve not gone anywhere. Though we’ve had an awful lot of traffic. Pleased to report some cool stats.

Average requests per user — 9.7
Avg # of pages served / day — 25,000
Bounce rate (we consider only 1 requesters a bounce): 27%

Our clearance time — and yes we use ngrok I am a big fan of theirs, hope there’s no ngrok haters here. We have our reasons — but :point_down:

These are our clearance times. Bear in mind we’re loading like, web pages. Mutating them. Parsing them. Returning them. And 90% of them we return in under 1 second.

This includes PDF’s up to 5 MB, websites, API calls — you name it.

Did we have a little downtime the other day? Yes. But hey, nobody’s perfect. We’ll get better :saluting_face:

Thought this was pretty cool today — also wanted to share . It’s as if GPT wakes from a coma.