|
1 | 1 | # Unreleased |
2 | 2 | * [changed] **Breaking Change**: `LiveModelFutures.connect` now returns `ListenableFuture<LiveSessionFutures>` instead of `ListenableFuture<LiveSession>`. |
3 | | - * **Action Required:** Remove any transformations from LiveSession object to LiveSessionFutures object. |
| 3 | + * **Action Required:** Remove any transformations from LiveSession object to LiveSessionFutures object. |
4 | 4 | * **Action Required:** Change type of variable handling `LiveModelFutures.connect` to `ListenableFuture<LiveSessionsFutures>` |
5 | 5 | * [changed] **Breaking Change**: Removed `UNSPECIFIED` value for enum class `ResponseModality` |
6 | 6 | * **Action Required:** Remove all references to `ResponseModality.UNSPECIFIED` |
|
13 | 13 | * [fixed] Fixed an issue with `LiveContentResponse` audio data not being present when the model was |
14 | 14 | interrupted or the turn completed. (#6870) |
15 | 15 | * [fixed] Fixed an issue with `LiveSession` not converting exceptions to `FirebaseVertexAIException`. (#6870) |
16 | | - |
17 | | - |
18 | | -# 16.3.0 |
19 | | -* [feature] Emits a warning when attempting to use an incompatible model with |
20 | | - `GenerativeModel` or `ImagenModel`. |
21 | | -* [changed] Added new exception type for quota exceeded scenarios. |
22 | | -* [feature] `CountTokenRequest` now includes `GenerationConfig` from the model. |
23 | | -* [feature] **Public Preview:** Added support for streaming input and output (including audio) using the [Gemini Live API](/docs/vertex-ai/live-api?platform=android) |
24 | | - **Note**: This feature is in Public Preview, which means that it is not subject to any SLA or deprecation policy and could change in backwards-incompatible ways. |
25 | | -* [changed] **Breaking Change**: `ImagenInlineImage.data` now returns the raw |
26 | | - image bytes (in JPEG or PNG format, as specified in |
27 | | - `ImagenInlineImage.mimeType`) instead of Base64-encoded data. (#6800) |
28 | | - * **Action Required:** Remove any Base64 decoding from your |
29 | | - `ImagenInlineImage.data` usage. |
30 | | - * The `asBitmap()` helper method is unaffected and requires no code changes. |
31 | | - |
32 | | -# 16.2.0 |
33 | | -* [fixed] Added support for new values sent by the server for `FinishReason` and `BlockReason`. |
34 | | -* [changed] Added support for modality-based token count. (#6658) |
35 | | -* [feature] Added support for generating images with Imagen models. |
36 | | - |
37 | | -# 16.1.0 |
38 | | -* [changed] Internal improvements to correctly handle empty model responses. |
39 | | - |
40 | | -# 16.0.2 |
41 | | -* [fixed] Improved error message when using an invalid location. (#6428) |
42 | | -* [fixed] Fixed issue where Firebase App Check error tokens were unintentionally missing from the requests. (#6409) |
43 | | -* [fixed] Clarified in the documentation that `Schema.integer` and `Schema.float` only provide hints to the model. (#6420) |
44 | | -* [fixed] Fixed issue were `Schema.double` set the format parameter in `Schema`. (#6432) |
45 | | - |
46 | | -# 16.0.1 |
47 | | -* [fixed] Fixed issue where authorization headers weren't correctly formatted and were ignored by the backend. (#6400) |
48 | | - |
49 | | -# 16.0.0 |
50 | | -* [feature] {{firebase_vertexai}} is now Generally Available (GA) and can be |
51 | | - used in production apps. |
52 | | - |
53 | | - Use the {{firebase_vertexai_sdk}} to call the {{gemini_api_vertexai_long}} |
54 | | - directly from your app. This client SDK is built specifically for use with |
55 | | - Android apps, offering security options against unauthorized clients |
56 | | - as well as integrations with other Firebase services. |
57 | | - |
58 | | - * If you're new to this library, visit the |
59 | | - [getting started guide](/docs/vertex-ai/get-started?platform=android). |
60 | | - |
61 | | - * If you were using the preview version of the library, visit the |
62 | | - [migration guide](/docs/vertex-ai/migrate-to-ga?platform=android) to learn |
63 | | - about some important updates. |
64 | | -* [changed] **Breaking Change**: Changed `functionCallingConfig` parameter type to be nullable in `ToolConfig`. (#6373) |
65 | | -* [changed] **Breaking Change**: Removed `functionResponse` accessor method from `GenerateContentResponse`. (#6373) |
66 | | -* [changed] **Breaking Change**: Migrated `FirebaseVertexAIException` from a sealed class to an abstract class, and marked constructors as internal. (#6368) |
67 | | -* [feature] Added support for `title` and `publicationDate` in citations. (#6309) |
68 | | -* [feature] Added support for `frequencyPenalty`, `presencePenalty`, and `HarmBlockMethod`. (#6309) |
69 | | -* [changed] **Breaking Change**: Introduced `Citations` class. Now `CitationMetadata` wraps that type. (#6276) |
70 | | -* [changed] **Breaking Change**: Reworked `Schema` declaration mechanism. (#6258) |
71 | | -* [changed] **Breaking Change**: Reworked function calling mechanism to use the new `Schema` format. Function calls no longer use native types, nor include references to the actual executable code. (#6258) |
72 | | -* [changed] **Breaking Change**: Made `totalBillableCharacters` field in `CountTokens` nullable and optional. (#6294) |
73 | | -* [changed] **Breaking Change**: Removed `UNKNOWN` option for the `HarmBlockThreshold` enum. (#6294) |
74 | | -* [changed] **Breaking Change**: Removed `UNSPECIFIED` option for the `HarmBlockThreshold`, `HarmProbability`, `HarmSeverity`, and `BlockReason` enums. (#6294) |
75 | | -* [changed] **Breaking Change**: Renamed `BlockThreshold` as `HarmBlockThreshold`. (#6262) |
76 | | -* [changed] **Breaking Change**: Renamed all types and methods starting with `blob` to start with `inlineData`. (#6309) |
77 | | -* [changed] **Breaking Change**: Changed the order of arguments in `InlineDataPart` to match `ImagePart`. (#6340) |
78 | | -* [changed] **Breaking Change**: Changed `RequestOption` to accept only `long` timeout values. (#6289) |
79 | | -* [changed] **Breaking Change**: Moved `requestOptions` to the last positional argument in the `generativeModel` argument list. (#6292) |
80 | | -* [changed] **Breaking Change**: Replaced sealed classes with abstract classes for `StringFormat`. (#6334) |
81 | | -* [changed] **Breaking Change**: Refactored enum classes to be normal classes. (#6340) |
82 | | -* [changed] **Breaking Change**: Marked `GenerativeModel` properties as private. (#6309) |
83 | | -* [changed] **Breaking Change**: Changed `method` parameter type to be nullable in `SafetySettings`. (#6379) |
84 | | - |
85 | | -# 16.0.0-beta05 |
86 | | -* [changed] Merged core networking code into VertexAI from a separate library |
87 | | -* [feature] added support for `responseSchema` in `GenerationConfig`. |
88 | | - |
89 | | -# 16.0.0-beta03 |
90 | | -* [changed] Breaking Change: changed `Schema.int` to return 32 bit integers instead of 64 bit (long). |
91 | | -* [changed] Added `Schema.long` to return 64-bit integer numbers. |
92 | | -* [changed] Added `Schema.double` to handle floating point numbers. |
93 | | -* [changed] Marked `Schema.num` as deprecated, prefer using `Schema.double`. |
94 | | -* [fixed] Fixed an issue with decoding JSON literals (#6028). |
95 | | - |
96 | | -# 16.0.0-beta01 |
97 | | -* [feature] Added support for `responseMimeType` in `GenerationConfig`. |
98 | | -* [changed] Renamed `GoogleGenerativeAIException` to `FirebaseVertexAIException`. |
99 | | -* [changed] Updated the KDocs for various classes and functions. |
100 | | - |
0 commit comments