Warp AI - AI directly integrated into the terminal

Warp AI - AI directly integrated into the terminal(warp.dev)

13 points by lord_sudo 3 years ago | 5 comments

szgupta 3 years ago |

Hi HN, I’m Suraj, an engineer at Warp (www.warp.dev) where we’re building a fast, Rust-based terminal for the 21st century. Today I am excited to share Warp AI with the community.

With this launch we have added AI to the core interface of the terminal. This integration allows you to ask Warp AI to explain command line errors and output, execute suggested fixes, walk you through complex CLI workflows, or write entire scripts for you.

For example, you can ask Warp AI to help you tail logs from a gcloud instance, generate new ssh certs, or run a suggested fix, e.g. ps aux | grep 'python' | awk '{print $2}' | xargs kill -9.

Warp AI also includes AI Command Search, which lets you generate any command using natural language — especially helpful for commands you know but can’t remember at the moment, saving you a trip to google or stack overflow.

All of this is built into the terminal interface itself which has benefits. There’s no need to run a separate CLI to access the AI interface. It lives within the terminal. You can run suggested commands directly from the AI response with the click of a button.

Here’s our official blog post announcing the feature: www.warp.dev/blog/introducing-warp-ai

Here’s a video of the feature in action: https://www.youtube.com/watch?v=1YZLPJ5X18k

A few notes on privacy and security: Telemetry in Warp — and Warp AI — is completely opt-in. The data we send is transparently documented [1] for reference. You can inspect all network requests made by Warp via our network log feature [2].

Also, when using Warp AI, no data from inputs is stored on Warp servers. The integration is built on OpenAI. Warp does not store any data, but rather proxies the requests.

It’s also important to call out: A login is required to use both Warp and Warp AI, similar to other cloud-based services.

Warp AI is in a free preview for developers on Mac, up to 100 requests per day.

Please try it out and let us know what you think!

[1] https://docs.warp.dev/getting-started/privacy#exhaustive-tel...

[2] https://docs.warp.dev/features/network-log

billconan 3 years ago | |

but if say I have a long shell script, or I have a long build job failed middle way, can it locate the relevant errors in the log?

szgupta 3 years ago | | |

Hey! Currently, when you share a block (grouped input and output) with Warp AI, we let the AI know the command that was ran and also the tail of the output. (and to streamline this even more, sharing a block with Warp AI is as simple as one-click!)

Often times, when these longer scripts / build jobs fail, they'll emit an error at the _end_, so sharing the tail should work in this case. But, you can always tweak (add / subtract) from what you want to share with the AI to make the query even more focused.

Does that address your use-case?