let’s focus on the ai agent. the purpose of this ai-agent is to help the user generate perfect images by using great prompts and choosing the best model and using autoimages unique prompt processing tools, it can ask the user what they are working on, it has in it’s context the current prompt textarea, the ai response should use tool calling to consistently return response: ” (required) and optional buttons: [{}] and newPrompt: ” buttons might just yes or no to if we are asking question or buttons can offer choices, newPrompt is a generated new prompt. we also want multiple prompts within 2 categories user and application, user prompts are obviously prompts triggered by a user containing the user prompts, application are types of prompts the client application may programmatically trigger and don’t have any user prompt. for example an application prompt which instructs the AI to say something like: “It looks like you are working on xyz, here is an improved version: abc123” based on the textarea value provided to initial prompt. where this gets complex is if the user prompted “let’s make it more photogenic” the ai is going to have to understand flux is the best photo model we have and it should recommend keywords like “photogenic 8k ultra-realism” and it could suggest using our Photogenic prompt helper checkbox. At some point we are going to need a new table. we are going to use OpenAI API and tool calling to define our responses. i would also like to include the last 2 user and system messages from the conversation history in OpenAI’s messages array remember the order is important.