NexusFlow Developer Documentation

One-stop integration for Qwen, DeepSeek, GLM, Kimi, HappyHorse, and more. Supports OpenAI, Anthropic Messages, and Gemini-compatible protocols with unified billing, keys, and monitoring.

Model Services

Unified model list, pricing, and capabilities

Three-Protocol Integration

OpenAI / Anthropic / Gemini compatible endpoints

Monitoring & Evaluation

Monitoring dashboard, error codes, and performance metrics

High-Concurrency Readiness

Rate limits, queues, task pipelines, and capacity scaling

Three Compatible Protocols

OpenAI-compatible

/v1/chat/completions

Anthropic Messages

/v1/messages

Suitable for reusing Anthropic SDK, Claude Code-style clients, and Messages request format.

Gemini-compatible

/v1beta/models/{model}:generateContent

Suitable for migrating from existing Gemini SDK or GenerateContent HTTP calls.

Quick Start

From your first chat request to async task integration

Model Overview

Browse model services, topic pages, and pricing

API Reference

OpenAI, Anthropic, and Gemini compatible protocols

Monitoring & Rate Limits

Rate limits, queues, and observability for production traffic

Popular Models

HappyHorse 1.0New

Alibaba . Video generation topic page and task pipeline integration

View Details

Qwen3 MaxRecommended

Alibaba Cloud . Flagship reasoning and complex task handling

View Details

Qwen3.5 PlusPopular

Alibaba Cloud . Best value, balanced and efficient

View Details

Qwen3.5 Max

Alibaba Cloud . Hundred-billion parameter, long-text expert

View Details

DeepSeek R1

DeepSeek . Open-source reasoning, code expertise

View Details

API Endpoints

POST/v1/chat/completionsChat Completions POST/v1/messagesAnthropic Messages Compatible POST/v1beta/models/{model}:generateContentGemini GenerateContent Compatible POST/v1/embeddingsText Embeddings POST/v1/tasksImage/Video Async Task Submission GET/v1/tasks/:idAsync Task Polling

Production Traffic Tips

Chat via Synchronous API

Chat and reasoning models should prefer `/v1/chat/completions` to reduce unnecessary polling complexity.

Multimedia via Async Tasks

Images and videos should use `/v1/tasks` uniformly, handling high latency and peak processing via task status.

Check Rate Limits & Monitoring Before Launch

Confirm peak request strategies on the rate limits, error codes, and monitoring pages to avoid discovering bottlenecks during traffic spikes.

Quick Example

from openai import OpenAI

client = OpenAI(
    api_key="sk-air-your-key",
    base_url="https://nexusflow.vip/v1",
)

response = client.chat.completions.create(
    model="qwen3.5-plus",
    messages=[
        {"role": "user", "content": "Hello!"}
    ],
)

print(response.choices[0].message.content)

Use the standard OpenAI SDK - just change the base_url to connect to NexusFlow.View full tutorial →