OpenAI has just announced a few new updates to its Assistants API, which is an API that makes it possible for developers to build AI assistants into their applications.
The API now includes a file search tool that allows up to 10,000 files per assistant, and can enable developers to integrate knowledge retrieval into their assistants. It works with OpenAI’s vector store objects to automate file parsing, chunking, and embedding.
The company also introduced new controls for setting the maximum input and output token so that developers can more directly limit costs. They can also now select how many recent messages will be used for context truncation.
It also introduced a new Tool Choice option that forces the model to call a specified tool during a run, such as file search, code interpreter, or another function.
Other updates in this release include more model configuration settings, new streaming and polling helpers, support for fine-tuned models, and more.
More information about these new additions can be found in the release notes.
