googleapis / googleapis/java-genai
Support POJO/Java Class for automatic responseSchema generation
- Dominant language
- Java
- Stars
- 397
- Forks
- 125
- Avg merge
- 2d 8h
- Merged PRs (30d)
- 41
Description
**Is your feature request related to a problem? Please describe.**
I am frustrated when I have to manually define the `responseSchema` using complex builders, even though I already have well-structured DTO (Data Transfer Object) classes in my Java project.
Currently, the Java SDK requires developers to manually construct a `Schema` object. This leads to redundant work and high maintenance costs, as any change in the Java class must be manually reflected in the `Schema` builder. For example, in a project with many complex DTOs, this manual mapping is highly error-prone and degrades the Developer Experience (DX).
**Describe the solution you'd like**
I would like the `GenerateContentConfig.Builder` to support an overloaded `responseSchema(Class clazz)` method. This method should automatically introspect the provided Java class (POJO/Record) and generate the corresponding `Schema` object.
**Desired Usage:**
```java
// Instead of manual Schema building
GenerateContentConfig config = GenerateContentConfig.builder()
.responseMimeType("application/json")
.responseSchema(MyResponseDto.class) // Passing the DTO class directly
.build();
```
The SDK should ideally use reflection or existing Jackson integration to map Java types (String, Integer, List, etc.) to the Gemini `Type` system.
**Describe alternatives you've considered**
I have considered the **Python Gen AI SDK**, which already provides this functionality through **Pydantic models**. Python developers can simply pass a class to `response_schema`, and the SDK handles the rest.
Currently, in Java, I have to write a custom utility using `jsonschema-generator` and map it back to the SDK's `Schema` object. This feels like a feature that should be natively supported to ensure "Feature Parity" across different language SDKs.
**Additional context**
Since the SDK already utilizes **Jackson** (`@JsonProperty`), implementing this wouldn't necessarily require heavy new dependencies. The implementation could respect Jackson annotations:
* `@JsonPropertyDescription`: To populate the `description` field in the schema.
* `@JsonProperty`: To handle custom field naming.
This enhancement would make the Java SDK feel much more "idiomatic" and productive for enterprise-level backend development.
Contributor guide
Assessment
This issue has not been assessed yet.