gpt

Call GPT generation and OpenAI text embedding models through RunAPI using OpenAI-compatible clients. Use for chat, Responses, embeddings, streaming, tools, vision, or an existing compatibility client that needs the conditional reference.

RunAPI@runapi-ai

Install

openclaw skills install @runapi-ai/runapi-gpt

GPT on RunAPI

Use OpenAI-compatible clients at https://runapi.ai/v1 as the primary protocol.

Primary protocol recipe

Authenticate

Set OPENAI_API_KEY to a RunAPI API key and OPENAI_BASE_URL to https://runapi.ai/v1.

Send request

python

from openai import OpenAI
client = OpenAI(api_key="YOUR_RUNAPI_TOKEN", base_url="https://runapi.ai/v1")
response = client.responses.create(model="gpt-5.5", input="Summarize this review.")
print(response.output_text)
print(response.usage)

Use client.chat.completions.create for gpt-5.2, which does not accept Responses. Use Responses for gpt-5.*-pro. Use client.embeddings.create only with the three embedding model IDs below. For streaming Responses, set stream=True and consume through one usage-bearing response.completed followed by [DONE].

For gpt-5.6-luna, gpt-5.6-sol, and gpt-5.6-terra, start with text input, sync or SSE transport, parameterized custom functions, and automatic prompt caching. Add reasoning controls, multimodal input, hosted tools, or continuation state only when the current RunAPI contract verifies that shape.

Verify result

Responses require final output, response.completed, and usage. Chat requires final assistant content, finish_reason, and usage. Embeddings require one numeric vector per input and response usage.

Stop boundaries

Correct a rejected shape once using the current error and endpoint contract. Retry transport once only before any response or Usage and when replay is safe. Record a terminal error and stop without changing model, endpoint, or protocol. Keep embedding models on /v1/embeddings.

Compatibility protocols

Load compatibility protocols only when an existing client requires Anthropic Messages or Gemini contents.

Live audio

gpt-live-1 is available only on the dedicated Live transport. Use WebRTC with POST /v1/live/webrtc and a client SDP offer, or WebSocket with GET /v1/live/websocket and a first session.start frame. Live sessions may delegate backend work through delegation.type: responses or return function work through delegation.type: client. Responses delegation accepts hosted tools on both transports and function tools on WebSocket: delegated events arrive wrapped in response.event, a function call arrives as a nested response.output_item.done, and the client answers with response.item.create (function_call_output) followed by response.create. Chat Completions and Responses calls do not accept this model.

Supported models

Model ID	Use when
`gpt-5.5`	Latest general model
`gpt-5.4`	Production default
`gpt-5.4-mini`	Cost-optimized
`gpt-5.4-nano`	Smallest, fastest
`gpt-5.2`	Cost-effective
`gpt-4.1`	Non-reasoning chat; image input on Chat Completions only
`gpt-4.1-mini`	Lower-cost GPT-4.1; image input on Chat Completions only
`gpt-4o-mini`	—
`gpt-5.6-luna`	Text and parameterized functions in the consistently available Responses subset
`gpt-5.6-sol`	Text and parameterized functions in the consistently available Responses subset
`gpt-5.6-terra`	Text and parameterized functions in the consistently available Responses subset
`gpt-6-astra`	—
`codex-auto-review`	—
`gpt-5-pro`	Text-only analysis and coding with the Responses API; other input modalities await billing review
`gpt-6-sol`	—
`gpt-6.1-sol`	—
`gpt-6-luna`	—
`text-embedding-3-large`	High-capacity vectors
`text-embedding-3-small`	Efficient vectors
`text-embedding-ada-002`	Legacy-compatible vectors

gpt