Releases: langchain-ai/langchain-google
libs/vertexai/v3.1.0
Changes since libs/vertexai/v3.0.3
feat(vertexai): support model profiles (#1374)
fix(vertexai): fix "Object of type 'FieldInfo' is not JSON serializable" across all schema generation (#1317)
fix(vertexai): mime type parsing from image URLs with params (#1377)
fix(vertexai): preserve cache_control during ToolMessage → tool_result conversion (#1369)
fix(genai, vertexai): thought signature handling (#1354)
chore(vertexai): mark callbacks and chains as deprecated (#1349)
chore(vertexai): remove gemma mentions in module docstring (#1348)
style(vertexai): fix pyproject.toml formatting (#1347)
feat(vertexai): add default dimensions parameter to VertexAIEmbeddings (#1242)
docs(vertexai): fix json_schema docstring typo (#1329)
chore(vertexai): bump lock (#1322)
chore(vertexai): bump pyarrow constraint to <23.0.0 (#1314)
libs/genai/v3.2.0
libs/genai/v3.1.0
Changes since libs/genai/v3.0.3
feat(genai): support gemini 3 (#1365)
chore(genai): update URLs for image tests (#1363)
feat(genai): store thought signatures in additional_kwargs (#1358)
fix(genai, vertexai): thought signature handling (#1354)
chore(genai)!: delete Attributed Question Answer (AQA) service interfaces (#1343)
chore(genai): mark google_vector_store module deprecated (#1350)
fix(genai): delete dead functions (#1346)
docs(genai): add example for get_num_tokens (#1345)
chore(genai): clean up pyproject.toml (#1344)
refactor(genai): additional tests work (#1328)
refactor(genai): tests (#1327)
libs/community/v3.0.1
Changes since libs/community/v3.0.0
release(community): 3.0.1 (#1335)
chore(community): allow Python 3.14, update deps (#1334)
chore: fix links (#1311)
chore(vertexai,genai,community): update project URLs in pyproject.toml and README.md files (#1308)
fix(vertexai,community): bump google-cloud-storage constraint to <4.0.0 (#1303)
chore: cleanup (#1290)
docs(community): update docstrings for reference docs #1237 (#1255)
style: more refs styling work (#1272)
style: refs pass, model names (#1270)
libs/genai/v3.0.3
libs/vertexai/v3.0.3
- Fix import blocking I/O in async contexts #1267
- Bump
google-cloud-storageconstraint to <4.0.0 #1303 - Add
transportas an alias ofapi_transportand fix somebase_urlbehavior #1293
Changes since libs/vertexai/v3.0.2
chore: fix links (#1311)
release(vertexai): 3.0.3 (#1310)
chore(vertexai,genai,community): update project URLs in pyproject.toml and README.md files (#1308)
chore(genai,vertexai): nits (#1307)
fix(vertexai,community): bump google-cloud-storage constraint to <4.0.0 (#1303)
fix(vertexai): test robustness improvement (#1301)
fix: base_url behavior, and improve checks and docstrings for base_url and api_transport behavior (#1293)
fix(vertexai): resolve blocking I/O in async contexts (#1267)
chore: cleanup (#1290)
libs/genai/v3.0.2
- Add support for Gemini's improved structured output #1288
- Add validation for empty contents in Gemini API requests #1298
- Fix blocking operation on import #1292
- Fix
base_url#1293 - Fix embeddings
task_typeusage #1302
Changes since libs/genai/v3.0.1
chore: fix links (#1311)
release(genai): 3.0.2 (#1309)
feat(genai): improved structured output (#1288)
chore(vertexai,genai,community): update project URLs in pyproject.toml and README.md files (#1308)
chore(genai,vertexai): nits (#1307)
revert: "extract and include metadata" (#1306)
fix(genai): resolve blocking I/O in async contexts (#1264)
fix(genai): Add validation for empty contents in Gemini API requests (#1298)
refactor(genai): update thinking model test to use parameterized inputs (#1219)
fix(genai): use task_type_to_use variable in embed_query methods (#1302)
fix: base_url behavior, and improve checks and docstrings for base_url and api_transport behavior (#1293)
docs(genai): enhance Jupyter notebook example for image generation (#1300)
test(genai): enhance assertions for structured output in chat model tests (#1295)
docs(genai): update invoke docstring to clarify code_execution parameter (#1296)
fix(genai): move importlib.metadata.version into module scope (#1292)
chore: cleanup (#1290)
libs/vertexai/v3.0.2
- Fix issue where thinking params were not being surfaced in traces
Changes since libs/vertexai/v3.0.1
release(vertex): 3.0.2 (#1284)
feat: Add base_url support (#1278)
test(vertex): use new claude model (#1273)
fix(vertex): make thinking params available for langsmith tracing (#1276)
test: fix fetching (#1281)
chore: bump langchain-tests (#1277)
style: more refs styling work (#1272)
style: refs pass, model names (#1270)
libs/genai/v3.0.1
Adds new functionality for video processing optimization and API endpoint customization, along with important bug fixes for thinking.
MediaResolutionparameter support (#1259) - Addedmedia_resolutionconfiguration toChatGoogleGenerativeAIfor optimizing video understanding. SettingMediaResolution.MEDIA_RESOLUTION_LOWcan reduce input token usage by 3x+ for video processing tasks.- Base URL customization (#1278) - Added
base_urlparameter as an alias for configuring custom API endpoints. This parameter maps toclient_options['api_endpoint']. - Extended thinking signature handling (#1269) - Fixed thought signature extraction during v1 content translation. The conversion function now properly decodes and preserves base64-encoded thought signatures from text blocks.
Changes since libs/genai/v3.0.0
release(genai): 3.0.1 (#1283)
feat: Add base_url support (#1278)
test: fix fetching (#1281)
chore: bump langchain-tests (#1277)
style: more refs styling work (#1272)
style: refs pass, model names (#1270)
fix(genai): catch thought signatures in v1 translation (#1269)
feat(genai): MediaResolution parameter support (#1259)