Investigate workarounds for OpenAI's `content_filter`

[inFocus7]

Sometimes when submitting requests to ChatGPT or DALL-E, the request is denied to to breaking the content_filter rules. This has previously happened when trying to generate "scary rooms with a sense of dread".

We should look into possible workarounds to this. Some thoughts:

1. We could retry if this error occurs.. but we need to look at the costs to re-submit a prompt. there would also need to be a limit on retries.
2. Support running local instances of StableDiffusion (and local LLM(s)) do not be restricted.
   • An issue + milestone exists for StableDiffusion support Investigate (local) StableDiffusion support #7

• relevant code