Update openai dep and add support for multimodal content

[transitive-bullshit]

• updates openai dep to latest (these changes seem mostly trivial)
• updates typescript dep to latest
• changes ChatMessage.content to support string | ChatMessageContentPart[] | null
  • this is going to have a decently large effect on downstream types, since we pretty much assume content is a string everywhere
  • but it's also necessary to support multimodal messages. at this time, only text and image_url types, but audio is also coming soon
  • here's an example where I had to cast messages to any to get image messages to work: https://github.com/transitive-bullshit/kindle-ai-export/blob/main/src/transcribe-book-content.ts#L50-L66

See https://platform.openai.com/docs/api-reference/chat/create for more details.

[rileytomasek]

i'll have to test this locally with dexter and the internal dexa code. i'm not sure what we'll want to do if it causes issues because we don't have time to update that code right now and don't need the non-text functionality.

also cc @cfortuner because this may conflict with the anthropic efforts that we also don't want to block for functionality we don't need.

[rileytomasek]

@transitive-bullshit what do you think of making a new type+method specifically for supporting mutimedia chat messages? it would be nice to have, but i dont think it's worth the hassle downstream for a ton of code that doesn't care about it.

[transitive-bullshit]

I think maintaining as much of a 1:1 w OpenAI makes sense. Otherwise they'll just drift apart as OpenAI goes more and more multimodal.

eg see https://platform.openai.com/docs/guides/audio which makes even more changes to the chat completions API to support multiple modalities.