Presets trailing new lines in prompt prefix are incorrect and depending on the AI provider (same behavior in firefox and chrome)
For example my gemini prefix has a single trailing newline char \n but two are inserted after the system prompt,
<system>You're a helpful coding assistant, follow these rules when answering:`talk like a pirate`</system>
first line
NOTE there also is an additional space before the chat content first line
Example gpt chat window:
<system>You're a helpful coding assistant, follow these rules when answering:`talk like a pirate`</system>
first line
mostly correct, but an additonal space before the chat content (probably intentional)
Example qwen response (identical to gpt)
Example Claude:
<system>You're a helpful coding assistant, follow these rules when answering:`talk like a pirate`</system> first line
ignores the new line all together.
Why I want this fixe.
I want to be able to read back my chats in the webchat UI, for this I want to have a newline between the prompt and the chat context.
Maybe replace the default space that is used to separate the prompt prefix from the chat content with a newline char \n still a single character and makes reading back the chat easier.
Settings.json:
"codeWebChat.chatPresetsForNoContext": [
{
"name": "Gemini pro",
"chatbot": "Gemini",
"promptPrefix": "<system>You're a helpful coding assistant, follow these rules when answering:`talk like a pirate`</system>\n",
"model": "pro"
},
{
"name": "ChatGPT thinking",
"chatbot": "ChatGPT",
"promptPrefix": "<system>You're a helpful coding assistant, follow these rules when answering:`talk like a pirate`</system>\n",
"options": [
"thinking"
]
},
{
"name": "Qwen",
"chatbot": "Qwen",
"promptPrefix": "<system>You're a helpful coding assistant, follow these rules when answering:`talk like a pirate`</system>\n",
"options": [
"thinking"
]
},
{
"name": "Claude",
"chatbot": "Claude",
"promptPrefix": "<system>You're a helpful coding assistant, follow these rules when answering:`talk like a pirate`</system>\n",
"model": "sonnet-4-5"
}
],
I'm using a windows 11 PC with vscodium
Presets trailing new lines in prompt prefix are incorrect and depending on the AI provider (same behavior in firefox and chrome)
For example my gemini prefix has a single trailing newline char
\nbut two are inserted after the system prompt,NOTE there also is an additional space before the chat content
first lineExample gpt chat window:
mostly correct, but an additonal space before the chat content (probably intentional)
Example qwen response (identical to gpt)
Example Claude:
ignores the new line all together.
Why I want this fixe.
I want to be able to read back my chats in the webchat UI, for this I want to have a newline between the prompt and the chat context.
Maybe replace the default space
that is used to separate the prompt prefix from the chat content with a newline char\nstill a single character and makes reading back the chat easier.Settings.json:
I'm using a windows 11 PC with vscodium