OpenAI Enhances Assistants API with Advanced File Management and Cost Control Features

OpenAI has upgraded its Assistants API with several new features aimed at enhancing the capabilities of the generative AI aides within applications powered by its models. The API now supports a file search tool that allows developers to manage up to 10,000 files per assistant for better knowledge retrieval through automated file parsing, chunking, and embedding using OpenAI’s vector store objects.

Assistants API

Support for tool choice lets you specify whether to use file search, code interpreter, or a particular function in a given run—increasing precision in your assistant’s operations.https://t.co/GfZTi7Z4YO pic.twitter.com/epaGW3WZr5

— OpenAI Developers (@OpenAIDevs) April 17, 2024

As part of a cost control measure, OpenAI also introduced controls for setting maximum input and output token limits. Developers can also now adjust the context truncation by selecting how many recent messages the assistant considers. Another major enhancement in the update is the introduction of Tool Choice, which mandates when and how the AI assistant utilizes specific tools, such as file search or a code interpreter, during its operation. There are also expanded model configuration settings, new aids for streaming and polling, support for fine-tuned models, and several other improvements, all designed to provide more flexibility and power to developers using the Assistants API.

OpenAI first debuted the Assistants API at its developer conference last year. The idea was to offer developers a way of making a kind of agent within an app that can follow specific instructions and use both OpenAI’s tools and proprietary knowledge bases. The company pitched it as a way to augment applications with everything from coding to vacation planning help leveraging OpenAI’s generative AI models. The API employs Code Interpreter, a feature that can carry out tasks in Python code within a controlled environment, enabling it to generate visuals and process large data files.

“With the new file search tool, you can quickly integrate knowledge retrieval, now allowing up to 10,000 files per assistant. It works with our new vector store objects for automated file parsing, chunking, and embedding,” OpenAI explained in a post on X. “Support for tool choice lets you specify whether to use file search, code interpreter, or a particular function in a given run—increasing precision in your assistant’s operations.”

Follow @voicebotaiFollow @erichschwartz

OpenAI Enhances GPT-4 Turbo Model for API and ChatGPT, Including Visual Input

OpenAI Enhances Generative AI Model Fine-Tuning and Custom Model Services

OpenAI Showcases New Generative AI Models and Lower API Prices, Cures GPT-4 ‘Laziness’