Thought I’d try using an action with my pickaxe, just the dall-e image generator. However, my pickaxe uses Claude API, while Actions (obviously for Dall-e) use openai API. My pickaxe attempted to create the image at the appropriate point, but then said it couldn’t generate images as it’s Claude.
Besides using openai API for my pickaxe (I’m not doing that), any suggestions to get around this?
You can solve this issue with prompting. Actions are not natively part of these LLMs. It’s a special thing Pickaxe has created for them. So sometimes you just need to explain to the LLM that it has the ability to do certain stuff.
Anthropic has been updating Claude, so its behavior might have changed.
Try adding a line like this to your prompt: “You can also generate images using the generate_dalle_3_image_own_key function. If the user asks for an image use the function.”
You can also use our default image generation (which uses Flux). It works well with Claude.