OpenAI SDK

If you have existing code built on the OpenAI SDK, point it at Pipevideo to access video generation. Pipevideo supports both OpenAI API shapes:

OpenAI method	Pipevideo endpoint	Best for
`client.responses.create`	`POST /v1/responses`	New integrations (recommended)
`client.chat.completions.create`	`POST /v1/chat/completions`	Existing chat-completions clients

Responses API (recommended)

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.pipevideo.co/v1",
  apiKey: process.env.PIPEVIDEO_API_KEY,
  defaultHeaders: {
    "HTTP-Referer": "https://myapp.com",
    "X-Pipevideo-Title": "My Video App",
  },
});

const response = await client.responses.create({
  model: "moonshotai/kimi-k2.5",
  input: "Animated logo with pulsing circles",
  webhook_url: "https://example.com/webhooks/pipevideo",
});

console.log(response.id, response.status);

Poll for the result:

const result = await client.responses.retrieve(response.id);
console.log(result.output_video_url);

Chat Completions API

Use this if your integration already calls chat.completions.create:

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.pipevideo.co/v1",
  apiKey: process.env.PIPEVIDEO_API_KEY,
});

const completion = await client.chat.completions.create({
  model: "moonshotai/kimi-k2.5",
  messages: [
    { role: "user", content: "Animated logo with pulsing circles" },
  ],
  webhook_url: "https://example.com/webhooks/pipevideo",
});

console.log(completion.id, completion.status);

Poll for the result:

const result = await client.chat.completions.retrieve(completion.id);
console.log(result.choices[0]?.message.video_url?.url);

Python

Responses API:

from openai import OpenAI

client = OpenAI(
    base_url="https://api.pipevideo.co/v1",
    api_key="pv_your_api_key",
)

response = client.responses.create(
    model="moonshotai/kimi-k2.5",
    input="Animated logo with pulsing circles",
    extra_headers={
        "HTTP-Referer": "https://myapp.com",
        "X-Pipevideo-Title": "My Video App",
    },
)

print(response.id, response.status)

Chat Completions API:

completion = client.chat.completions.create(
    model="moonshotai/kimi-k2.5",
    messages=[
        {"role": "user", "content": "Animated logo with pulsing circles"},
    ],
)

print(completion.id, completion.status)

Differences from OpenAI

OpenAI	Pipevideo (Responses)	Pipevideo (Chat Completions)
Text in `output_text`	Video URL in `output_video_url`	Video URL in `choices[0].message.video_url`
Async — poll or pass `webhook_url`	Async — poll or use webhooks	Async — poll or use webhooks
`model` is a chat model	`model` is an orchestration LLM id	`model` is an orchestration LLM id
Per-token or per-request pricing	Token-based orchestration billing	Token-based orchestration billing

Recommended alternative

For the best experience with polling, waiting, and typed video responses, use the official TypeScript SDK (npm install pipevideo).

App attribution — optional headers for rankings
Responses API — full HTTP reference
Chat completions API — legacy-compatible HTTP reference

Verification

The API includes integration tests that run the official OpenAI SDK and Vercel AI SDK (@ai-sdk/openai + ai) against the real oRPC handler with a mocked backend:

cd apps/api && npm run test:integration