openai-api

What it does

Outlines how to interact with OpenAI stateless APIs such as Chat Completions, Embeddings, Images, Audio, and related features. Describes models in the GPT-5 and GPT-4o families, including streaming, function calling, structured outputs, and vision capabilities. Covers how to request and handle responses, including messages structure, streaming, tool calls, and JSON schemas for structured outputs. Includes guidance on environment setup and installation for the openai package.

How it works

Provides concrete code examples for:

• Creating chat completions with models (e.g., gpt-5, gpt-5.2, o3) using OpenAI SDK or fetch, including streaming and function calling workflows.
• Defining tools for function calling and handling tool_calls, then re-invoking the API with tool results until no tool_calls remain.
• Using Structured Outputs via JSON Schema to enforce response formats.
• Handling Vision (GPT-4o) with image input either via URL or base64 data.
• Accessing Embeddings API with models like text-embedding-3-small or -large, including batch processing and custom dimensions.
• Interacting with Images API for DALL-E 3 generation and image edits, including multipart/form-data transfer.

When to use it

Use when implementing GPT-5 chat, streaming, function calling, embeddings for RAG, or troubleshooting rate limits (429) and API errors, as well as TypeScript issues and model name errors.

What it can touch

• Node.js SDK (openai package) and fetch approaches are demonstrated.
• Tools for function calling are defined as JSON schemas and invoked via openai.chat.completions.create with a tools array.
• JSON schemas for Structured Outputs are demonstrated through response_format with json_schema.
• Image and audio endpoints (images/generations, images/edits, whisper) are described.

Caveats

• The guide references specific models and parameters (e.g., gpt-5.2, reasoning_effort values) and notes limitations (e.g., GPT-5 not supporting temperature, top_p). It does not provide guarantees of outcomes and relies on OpenAI API behavior as described.
• Some sections show examples that depend on environment setup and network access; ensure correct API keys and environment variables.