:CopilotChatModels / :CopilotChatAgents don't change the actual AI engine

[tecfu]

Steps To Reproduce:

1. Add and authorize the Github copilot extension "models"

2. Change the AI Agent and Model via command line as shown below:

:CopilotChatAgents > models
:CopiliotChatModels > claude-3.7-sonnet 

3. Type a message into the Copilot Chat window to verify that your AI engine is now changed:

What AI engine are you based on?

4. Observe Copilot Chat respond that it is using OpenAI Codex

Expected Result:

Copilot Chat should indicate that it is using claude-3.7-sonnet

[deathbeam]

asking LLMs what model are they was never reliable way to verify it. You can turn on debug and see chat the api responses and model field to verify your model config:

[CopilotChat] data: {"choices":[{"index":0,"delta":{"content":" with your programming project today?","role":"assistant"}}],"created":1741672965,"id":"03bf4ada-a952-49e2-bdc6-adf1620011cd","model":"claude-3.7-sonnet"}

But otherwise most AI interfaces just hardcode the actual model answer in default system prompt. For example claude: https://docs.anthropic.com/en/release-notes/system-prompts#feb-24th-2025

I can technically hardcode it as well but idk if it serves any point really other than wasting tokens