Conversation
PiperOrigin-RevId: 886316760
PiperOrigin-RevId: 886659065
PiperOrigin-RevId: 886722180
PiperOrigin-RevId: 886723823
PiperOrigin-RevId: 886725038
…p types PiperOrigin-RevId: 886727168
…main PiperOrigin-RevId: 886751193
PiperOrigin-RevId: 886764077
Merge google#1067 🤖 I have created a release *beep* *boop* --- ### Updating meta-information for bleeding-edge SNAPSHOT release. --- This PR was generated with [Release Please](https://github.com/googleapis/release-please). See [documentation](https://github.com/googleapis/release-please#release-please). COPYBARA_INTEGRATE_REVIEW=google#1067 from google:release-please--branches--main 0b85f5b PiperOrigin-RevId: 886764460
This CL enables the simultaneous use of `output_schema` (structured output) and `tools` for models that do not natively support both features at once (specifically Gemini 1.x and 2.x on Vertex AI). ### Core Logic The CL implements a workaround for models with this limitation: 1. **Synthetic Tooling**: Instead of passing the `output_schema` directly to the model's configuration, it introduces a synthetic tool called `set_model_response`. 2. **Schema Injection**: The parameters of this tool are set to the requested `output_schema`. 3. **Instruction Prompting**: System instructions are appended, directing the model to provide its final response using this specific tool in the required format. 4. **Response Interception**: The `BaseLlmFlow` is updated to check if `set_model_response` was called. If so, it extracts the JSON arguments and converts them into a standard model response event. ### Key Changes * **`OutputSchema.java` (New)**: A new `RequestProcessor` that detects when the workaround is needed, adds the `SetModelResponseTool`, and provides utilities for extracting the structured response. * **`SetModelResponseTool.java` (New)**: A marker tool that simply returns its input arguments, used to "capture" the structured output from the model. * **`ModelNameUtils.java`**: Added logic to identify Gemini 1.x and 2.x models and determine if they can handle native `output_schema` alongside tools. * **`BaseLlmFlow.java`**: Updated the flow logic to detect the synthetic tool response and generate the final output event. * **`Basic.java`**: Updated to prevent native `outputSchema` configuration when the workaround is active. * **`SingleFlow.java`**: Registered the new `OutputSchema` processor. PiperOrigin-RevId: 886769688
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
6 participants
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.
No description provided.