Common Tool Integration

Clients that support a custom OpenAI-compatible address usually only need a Base URL, an API key, and a model ID. The model catalog is based on the real-time result of /v1/models.

OpenAI-compatible clients cover the model catalog, text chat, image generation, video tasks, and text-to-speech APIs; the asset library APIs use the platform's own /api/v1/materials/... paths, so please send HTTP requests directly as described in Asset Library API Overview; clients are not responsible for this part.

General configuration

If the tool asks for a full URL, use https://locopalai.com/v1/chat/completions; if it asks for a Base URL, fill in up to /v1.

The platform may also show addresses such as https://us-api.locopalai.com for overseas models or specific network environments. Use the address currently shown in the console when integrating; do not put both domains into the same client configuration.

OpenAI JavaScript SDK

npm install openai

import OpenAI from 'openai'

const client = new OpenAI({
  apiKey: process.env.LOCOPAL_API_KEY,
  baseURL: 'https://locopalai.com/v1',
})

const models = await client.models.list()
const result = await client.chat.completions.create({
  model: models.data[0].id,
  messages: [{ role: 'user', content: '你好' }],
})
console.log(result.choices[0].message.content)

OpenAI Python SDK

pip install openai

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["LOCOPAL_API_KEY"],
    base_url="https://locopalai.com/v1",
)
models = client.models.list()
response = client.chat.completions.create(
    model=models.data[0].id,
    messages=[{"role": "user", "content": "你好"}],
)
print(response.choices[0].message.content)

Cherry Studio

1. Open the model service settings and add an OpenAI-compatible service.
2. For the API address, enter https://locopalai.com/v1.
3. For the API key, enter the key created in the console.
4. Click to fetch models, or manually enter a model ID returned by /v1/models.
5. After saving, start a new conversation to verify.

If the model list is empty, first use curl to verify /v1/models with the same key; if curl works, check whether the client is appending /v1 twice.

OpenCode

Configure an OpenAI-compatible Provider in opencode.json at the user or project level:

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "locopal": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "LocoPal Ai",
      "options": {
        "baseURL": "https://locopalai.com/v1",
        "apiKey": "{env:LOCOPAL_API_KEY}"
      },
      "models": {
        "你的模型ID": { "name": "你的模型ID" }
      }
    }
  },
  "model": "locopal/你的模型ID"
}

The configuration syntax may change with client versions; follow the current client's configuration specification. The key point is that the Base URL, API key, and model ID are consistent.

Trae and other IDEs

In "Settings → Models → Add Model", choose the OpenAI-compatible format:

1. When "Full URL" is off, enter https://locopalai.com/v1.
2. When "Full URL" is on, enter https://locopalai.com/v1/chat/completions.
3. Enter the API key.
4. Enter a currently available model ID.
5. Enable vision or tool calling options according to the model's real capabilities.

Postman / Apifox

Create three environment variables: base_url=https://locopalai.com, api_key=你的APIKey, and model=从 /v1/models 获取的模型 ID. Add the following to each request:

Authorization: Bearer {{api_key}}
Content-Type: application/json

For video tasks, first save the ID from the create response, then use GET to query {{base_url}}/v1/video/generations/{{task_id}}.

Server-side proxy recommendation

Browsers and mobile apps should not hold the platform API key directly. The recommended architecture is "web page or app → your business backend → LocoPal Ai". The business backend handles user authentication, parameter validation, request logging, task ID persistence, quota control, and result storage.

Integration self-check

1. GET /v1/models returns at least one model.
2. A text call returns choices.
3. Video creation returns a task ID.
4. Video queries return a status and do not create duplicates.
5. When the current group provides a text-to-speech model, POST /v1/audio/speech returns playable binary audio.
6. On 401, check the key; on 429, wait as instructed; on 5xx, save the RequestId and then decide whether it can be retried.