Provide a way to get `content` instead of `reasoning_content`

[dkotter]

If you use a model that supports reasoning, the response you'll get back contains both a reasoning_content part and content part. The former contains the full reasoning response while the latter contains the final response, which I think is what you normally are wanting.

Right now if I'm using a reasoning model and I call the generateText method, this returns a GenerativeAiResult object and then calls the toText method of that.

This method will return the first message part that contains text and when we first parse the API response, we set the reasoning_content piece as the first message part and so this is what gets returned. Same thing happens if you initially use generateTexts.

As far as I can see, there isn't a way, other than parsing the response myself, to force it to use the content piece instead of reasoning_content. I see in #52 we added a way to distinguish between content and thought, seems like it may be useful to use that within these generate methods to determine which message part we return. Curious for thoughts on that though or if there's a better way to achieve this.

[felixarntz]

@dkotter Great point. I'd consider this simply a bug. toText(), toTexts() as well as any other to*() methods in GenerativeAiResult should only return the data from a part that is the content channel, i.e. not a thought channel.

I think this would be the most reasonable approach to address it. Since it only is meant to return the first text from the first part, these methods are meant for simple-structure responses that typically consist of only a single part.

So returning only the content (vs the thought and content) makes most sense to me. If you want to get both, then you're properly dealing with multi-part content, i.e. custom processing is probably required.

[dkotter]

Opened #92 to make this change, though open to discussion on the best approach there