1.0.0-beta.4 (2024-08-29)

Features Added

• file_search tool definitions now have the ability to configure max_num_results via the matching, named inner options object Previously, only a stubbed { "type": "file_search" } was valid. Now, e.g.: { "type": "file_search", "file_search": { "max_num_results": 20 } }

  • Added a new property FileSearchToolDefinitionDetails fileSearch to FileSearchToolDefinition model.
  • Added new class FileSearchToolDefinitionDetails to represent the details of a file search tool.

• chunking_strategy is added as a new optional property when creating a vector store (either via the vector store creation operation or the helper when creating an assistant) -- this allows customization of the chunk size and overlap used when ingesting data. See the OpenAI reference for full details.

  • Added a new property VectorStoreChunkingStrategyRequest chunkingStrategy to VectorStoreOptions model.
  • Added a new property VectorStoreChunkingStrategyResponse chunkingStrategy to VectorStoreFile model.
  • Added new enum VectorStoreChunkingStrategyRequestType and VectorStoreChunkingStrategyResponseType to represent the chunking strategy type for vector stores.
  • Added new class VectorStoreChunkingStrategyRequest and VectorStoreChunkingStrategyResponse to represent the chunking strategy for vector stores.
  • Added new class VectorStoreAutoChunkingStrategyRequest and VectorStoreAutoChunkingStrategyResponse to represent the auto chunking strategy for vector stores.
  • Added new class VectorStoreStaticChunkingStrategyOptions, VectorStoreStaticChunkingStrategyRequest and VectorStoreStaticChunkingStrategyResponse to represent the static chunking strategy for vector stores.

Other Changes

Dependency Updates

• Upgraded azure-core to version 1.51.0.
• Upgraded azure-core-http-netty to version 1.15.3.