Copilot chat ignoring context even though it clearly is referenced

[Superskyyy]

• VS Code Version: 1.85.2
• plugin version: v1.155.679
• OS Version: Linux debian12
• Logs:

I'm sorry for the confusion, but as an AI, I don't have the ability to see or select code in your environment. Could you please copy and paste the code you want me to check into our conversation?

Steps to Reproduce:

1. Select a long cpp file with 510 lines.
2. Ask copilot to explain or do whatever, it will say it has no access to it.
3. Select a small portion of the file (e.g. 200lines)
4. Ask again, now it is willing to answer questions and can see code.

[tecunited]

I've been having the same issue for 3 days, it just suddenly mid work it suddenly stopped.

[Superskyyy]

I've been having the same issue for 3 days, it just suddenly mid work it suddenly stopped.

yeah, its very random

[datanetai]

Context size is only 3k which you can see by going output from vscode terminal.

2024-02-02 11:27:20.296 [info] [chat fetch] modelMaxPromptTokens 3072
2024-02-02 11:27:20.296 [info] [chat fetch] modelMaxResponseTokens 1498
2024-02-02 11:27:20.296 [info] [chat fetch] chat model gpt-4

It cannot process 500 lines of code! I want them to show at least error or something it is weird because it gives me random code and i do not know if it can see my code or not. Also 3k context is too small for paid product which switch to GPT3.5 mid answering anyway!

[Shade53]

I am getting this issue with only 200 lines of code being presented, also running it in debian12 like OP. Just randomly started happening today when it was working a couple chats before. Upon checking my file though I see the full 200 lines of code is around 11.9KB, so it may be hitting this "3k" limit?

[MikeImbar]

I am also having the same issue with ~400 lines of code.

[ShawnZuspaet]

this is happening all the time to me, feels like it got worse.