Build OpenAI requests for batch processing

Usage

oai_build_completions_request_list(
  inputs,
  endpointr_ids = NULL,
  model = .OAI_DEFAULT_MODEL,
  temperature = NULL,
  max_tokens = 500L,
  schema = NULL,
  system_prompt = NULL,
  max_retries = 5L,
  timeout = 30,
  key_name = "OPENAI_API_KEY",
  endpoint_url = "https://api.openai.com/v1/chat/completions"
)

Arguments

inputs: Character vector of text inputs
endpointr_ids: A vector of IDs which will persist through to responses
model: OpenAI model to use
temperature: Sampling temperature (0-2), included in the request only when non-NULL. The default NULL omits it, which reasoning models (the GPT-5 family, o-series) require - they only accept the default temperature.
max_tokens: Maximum tokens per response, sent as OpenAI's max_completion_tokens (max_tokens is deprecated and rejected by reasoning models)
schema: Optional JSON schema for structured output
system_prompt: Optional system prompt
max_retries: Integer; maximum retry attempts (default: 5)
timeout: Numeric; request timeout in seconds (default: 30)
key_name: Environment variable name for API key
endpoint_url: OpenAI API endpoint URL

Value

List of httr2 request objects