googleapis / googleapis/google-cloud-java

Add built-in support for exposing the response schema as a structured object map

Offen
#11,628 2 Kommentare 0 Reaktionen 0 zugewiesene Personen Auf GitHub ansehen
priority: p3 type: feature request
Vorherrschende Sprache
Java
Sterne
2.1k
Forks
1.2k
Ø Merge
1 T. 23 Std.
Gemergte PRs (30 T.)
154

Beschreibung

**Is your feature request related to a problem? Please describe.**
Currently, when using the Java Vertex AI SDK’s `generateContent()` (e.g. via `ContentMaker`), we only get back a raw JSON string. Client code must separately construct a JSON Schema, attach it in `GenerationConfig`, then manually parse the returned JSON—either via Jackson’s `ObjectMapper` or via Protobuf’s `JsonFormat`. This boilerplate is brittle, error-prone, and hides what fields the model will actually return.

**Describe the solution you’d like**

1. Expose the active response schema programmatically. For example:

```java
Schema schema = model.getResponseSchema();
```

or as metadata on `GenerateContentResponse`.
2. Provide a one-step parser that binds the response into a POJO, Map, or Protobuf message, e.g.

```java
MyResult result = response.getTypedContent(MyResult.class);
```

under the hood handling all JSON Schema validation and mapping.

**Describe alternatives you’ve considered**

* Continuing to manually wire up JSON Schema definitions + `GenerationConfig`.
* Writing custom helper utilities in each codebase to wrap `ObjectMapper` or `JsonFormat`.
* Using generic `Map` parsing and then casting fields at runtime.

**Additional context**

* Sample desired usage:

```java
Schema schema = model.getResponseSchema();
MyResponse resp = model.generateContent(prompt, config)
.getTypedContent(MyResponse.class);
```
* This would eliminate repetitive boilerplate, reduce parsing errors, and let developers discover response fields at compile time.

Beitragsleitfaden

Beitragsleitfaden öffnen

Bewertung

Dieses Issue wurde noch nicht bewertet.

Neue Issues direkt in Ihr Postfach

Eine kurze Übersicht über anfängerfreundliche GitHub-Issues.