OpenAI Adapter Capabilities

The OpenAI Adapter enables you to call large language models (LLMs) from integrations in Oracle Integration. The OpenAI APIs provide a simple REST interface to state-of-the-art LLM models for text generation, prompting, and function calling. Depending on the OpenAI API you use, you can call OpenAI, Anthropic, and Azure LLMs with the OpenAI Adapter.

The OpenAI Adapter provides the following capabilities:

• Supports the following APIs. Each API offers different capabilities:

  • The Responses API lets you call OpenAI LLMs. See Responses.
  • The Chat Completions API lets you call OpenAI, Anthropic, and Azure LLMs. See Chat Completions.
  • The Upload file API lets you upload files to an OpenAI account to fine-tune models, use with certain endpoints (such as the files object), or use with tools such as assistants or function calling. See Upload file.
  • The List files API lets you obtain a list of all files you previously uploaded to an OpenAI account. This action enables you to manage your uploaded files, check their status, and reference them in other API requests (for example, for fine-tuning or assistants). See List files.

  You select the API to use when configuring the OpenAI Adapter in the Adapter Endpoint Configuration Wizard. See Basic Info Page.

• Simple and extended prompt-based interactions to define the structure of the input: You provide instructions, questions, or plain text to generate contextual and coherent AI responses from a prompt. The text response can be code, mathematical equations, structured JSON data, or human-like prose.

• Role and context definitions: The following roles influence how the model responds and behaves.

  • platform: Messages added by OpenAI. These roles carry the highest authority and override all other instructions.
  • developer: Input from the Oracle Integration application developer.
  • user: Input from end users, or a catch-all for data provided as input to the model.
  • assistant: Input sampled from the LLM model.
  • tool: Input generated by some program, such as code execution or an API call.

• Context maintenance: The previous response ID is used to support ongoing conversations and follow-up questions.

• Model selection and tuning: You select from available OpenAI models and fine-tune responses with parameters such as temperature.

• Flexible input options: Simple prompts or structured arrays containing role/content pairs are supported.

• Function calling support: OpenAI models can interface with code or external services.

• Use cases: A variety of use cases are supported, including the following:

  • Text analysis: Understand and extract insights from textual data (that is, analyze employee sentiment from appraisal or customer feedback from product reviews).
  • Text categorization: Automatically classify documents or resumes based on content and predefined rules.
  • Email drafting: Generate intelligent email responses for customer queries, HR communication, or marketing campaigns.
  • Translation: Translate customer support responses or documentation into multiple languages.
  • Summaries: Condense long documents or reports into digestible summaries for decision makers.

You can configure the OpenAI Adapter as an invoke connection in an integration in Oracle Integration. The OpenAI Adapter is one of many predefined adapters included with Oracle Integration. See the Adapters page in the Oracle Help Center.