How to add an image in the AI prompt?
How to add an image in the AI prompt?
Answers
-
Images and videos are not supported right now. What's your use-case?
-
It can be a screenshot of a webpage and telling AI to do something about it. Or a screenshot of an error. Like we can paste an image in the chat bot doe chatgpt or Perplexity.
-
I would love this too or a mechanism that will allow referencing files. It could function how AI/LLM CLIs and apps work or accept full paths, C:\Users\\Documents\Snagit\SNAG-xxx.png
-
If you're using Copilot, you can reference files right now via their full path. This is really useful for attaching other source code files, log files, JSON, XML, etc., and is what allows the following kind of instruction:
Write a parser for c:\temp\log.txt and summarize its content
However, I would be surprised if GitHub Copilot has unlocked image processing capability on their models via the SDK, given how much money they're bleeding right now. A quick test indicates that it will read the .png, but not analyze it other than for exif data.
-
On a slight tangent, are you people open to using Claude Code? It's arguably the gold standard in coding agents and offers high-value plans (something that's no longer the case with GitHub Copilot as of June 1).
I'm currently in the early stages of prototyping CC integration in LINQPad. It handles images just fine - Alt+V to paste them in. You'd get the same command-line experience but hosted inside LINQPad, and with Claude able to interact with the current script.
-
I> @JoeAlbahari said:
On a slight tangent, are you people open to using Claude Code? It's arguably the gold standard in coding agents and offers high-value plans (something that's no longer the case with GitHub Copilot as of June 1).
I'm currently in the early stages of prototyping CC integration in LINQPad. It handles images just fine - Alt+V to paste them in. You'd get the same command-line experience but hosted inside LINQPad, and with Claude able to interact with the current script.
Yes. I am already using CC with LINQPad as the harness with claude -p and Util.Cmd/Util.ReadLine. I have a UI dumped that's a simplifed terminal CLI. Having a dedicated tab or window for CC or Codex would be better integration. Now LINQPad is becoming an IDE. The Al Chat tab and the CC integration might benefit from a split view UI or lightweight docking system instead of being tied to a linq or cs file. I always wanted support for chats with txt/md files as input. I made custom tools for this when AI was added,
long before agents. The CC integration should include that and more naturally. I use copilot, claude and codex. Perhaps the integration could allow you to select the SDK, or could you make it extensible. -
Right now, the plan is for it to be at least loosely tied to the current tab in terms of editor integration, although it will be able to read files from anywhere in the filesystem (the same is true right now with LINQPad's Copilot integration). So you can ask it to analyze a txt/md file (or multiple files) just by providing the path to the files.
You'll also be able to transfer/fork chats to other scripts with the /resume command (there will also be a dedicated button for this). And the current "List all chats" feature will work for Claude chats, too, so you won't lose your chats if there are lot of tabs active.
The integration is a big undertaking - particularly the cross-platform terminal emulation and IPC - but so far it's progressing very well.
