OpenAI¶

Uses the Responses API, not chat completions. It is OpenAI's current surface and exposes reasoning effort and reasoning-token accounting the older shape does not.

Want the chat-completions dialect instead? Point openai-compat at https://api.openai.com/v1.

streaming structured output tool calls health discovery

Setup¶

client = ai.Client([
    ai.ProviderSettings.of("openai", api_key="env://OPENAI_API_KEY"),
])
result = client.generate(prompt, target="openai:gpt-5")

No extra required — this is raw httpx2.

Supported¶

Behavior	Support
Streaming	Native, typed events
Structured output	`json_schema` via `text.format`
Tools	Native
Reasoning	`reasoning.effort`, plus reasoning-token counts
Usage	Input, output, cached, reasoning tokens
Cost	Catalogued pricing

Reasoning¶

result = client.generate(prompt, target="openai:gpt-5", reasoning="high")
result.usage.reasoning_tokens

Effort levels pass straight through: minimal, low, medium, high.

Notes¶

System messages become the top-level instructions field.
The output-token parameter is max_output_tokens.
A response truncated by the token cap reports finish_reason == "length".

Provider options¶

provider_options={"openai": {"store": False, "service_tier": "flex"}}

Wire contract¶

For the exact request/response fields this adapter depends on, see contracts/openai.md.