New
google_generative_ai-v0.4.0
- Add support for parsing Vertex AI specific fields in
CountTokensResponse. - Add named constructors on
Schemafor each value type. - Add
GenerationConfig.responseMimeTypewhich supports setting'application/json'to force the model to reply with JSON parseable output. - Add
outputDimensionalityargument support forembedContentandbatchEmbedContent. - Add
Content.functionResponsesutility to reply to multiple function calls in parallel. - Breaking The
Partclass is no longersealed. Exhaustive switches over aPartinstance will need to add a wildcard case.