I am seeing that it is possible to format (both user, assistant, and system) messages this way
{
"role": "user",
"content": "Can you count up to 10?"
}
or this way
{
"role": "user",
"content": [
{
"type": "text",
"text": "Can you count up to 10?"
}
]
}
i.e. the difference is that you make a "content" a list of the dictionaries
second option gives more flexibility as it allows adding images that list example
{
"role": "user",
"content": [
{
"type": "text",
"text": "This is previous context blah blah blah. Now main question is: What is in this image (specifically the money in euros, generate EURO sign in the response)?"
},
{
"type": "image_url",
"image_url": {
"url": "https://d3dqioy2sca31t.cloudfront.net/Projects/cms/production/000/032/621/original/248bc404f13f59cb775c194cedcf08a6/article-tips-money-euros-in-hand.jpg"
}
}
]
},
So what is really the right way of formatting your messages (both without any files or with) according to OpenRouter team?