@tailored-ai/provider-openai

GPT and o-series models via chat completions, with the request shaping current OpenAI models require.

This plugin is the home of the openai provider id (#236 moved it out of core) — your existing providers.openai block keeps working. On older cores that still ship a minimal built-in, the plugin supersedes it; the one-line "Replacing existing entry" startup notice is expected there.

Why a plugin?

Reasoning models (o3, o4-mini, the gpt-5 family) reject temperature and the deprecated max_tokens parameter — the request shaping has to track OpenAI's API, which moves faster than core should. This plugin:

omits temperature for reasoning models (prefix heuristic, extendable via reasoningModels)
always uses max_completion_tokens
adds OpenAI-Organization / OpenAI-Project headers
passes ChatParams.extra through (e.g. reasoning_effort)
implements chatStream and listModels

Install

bash

tai plugin install @tailored-ai/provider-openai

Configure

yaml

providers:
  openai:
    apiKey: "${OPENAI_API_KEY}"
    defaultModel: "gpt-5-mini"

agent:
  defaultProvider: openai

Field	Required	Notes
`apiKey`	yes	Use `${OPENAI_API_KEY}` to read from the environment.
`defaultModel`	yes	Any chat-completions model id. `tai edit` lists them via `listModels`.
`organization`	no	`OpenAI-Organization` header.
`project`	no	`OpenAI-Project` header.
`reasoningModels`	no	Extra model-id prefixes treated as reasoning models.
`baseUrl`	no	Default `https://api.openai.com/v1`.