[WIP] Gemini Structured Outputs #116
Open
+193
−1
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Hey folks, this is a bit beyond my comfort zone, but wanted to share my WIP code for structured outputs with Gemini.
This code is working for me with a handful of somewhat complex schemas (I've explicitly used/ tested the
Object
,Array
,String
andNumber
schemas, with nesting), although I acknowledge I haven't attached any tests here (yet).Some observations that may need to be addressed more meaningfully:
ObjectSchema
defines anallowAdditionalProperties
key which is valid for OpenAI, but not for Gemini. As the schema itself isn't aware of theProvider
being chosen, without doing some injection/ facade/ whatever, it was easier to just strip this key out after thetoArray()
map. There may need to be a longer term decision made here based on how other providers treat their schema definitions, or perhaps a provider-specificmap()
ortoArray()
handler for theschema
.StructuredModeResolver
is a bit superfluous. Granted it's still a valid check on if a specific model supports structured mode, but I skipped this logic inclusion in my WIP pass here as all "current" Gemini models (excluding v1 Pro) support structured outputs.Thoughts on a path forward to formalizing this?