API Access Documentation

Connection Information

Connect applications only to the official Taida Nexus domain. The platform routing layer handles provider, region, and channel selection.

1. 1
   Create an API key

   Set model scope, credit limit, expiration, and IP allowlist for each project.

2. 2
   Configure the Base URL

   Use the unified platform address in your OpenAI-compatible SDK.

3. 3
   Select a Model ID

   Call GET /v1/models and use one of the returned model IDs.

Agents & IDE Integration

Taida Nexus supports the OpenAI-compatible protocol. Agents, IDE extensions, and orchestration frameworks that accept a custom Base URL can connect using three settings:

• Programming assistants and IDEs: in Cursor, Cline, Roo Code, Continue, or Aider, select the OpenAI Compatible provider and enter the three values above.
• Desktop & chat clients: Cherry Studio, NextChat, LobeChat, ChatBox, and similar clients use the same settings.
• Orchestration platforms & frameworks: Dify, LangChain, LlamaIndex, AutoGen, and custom OpenAI SDK agents can point base_url to the platform address.
• Function calling: The chat endpoint supports tools, tool_choice, streaming, and streamed tool calls.
• Model selection: Use auto when you do not want to choose a model. Billing records show the model that was actually used.

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.taidatech.ai/v1",
  apiKey: process.env.NEXUS_API_KEY,
});

// Streaming and function calling (tools / tool_choice) are supported.
const resp = await client.chat.completions.create({
  model: "auto",
  messages: [{ role: "user", content: "Read package.json and summarize the project." }],
  tools: myAgentTools,
});

Anthropic-compatible clients

The platform provides an Anthropic Messages-compatible endpoint at POST /v1/messages. Configure Claude Code and similar clients with these environment variables:

# Claude Code and other Anthropic-compatible clients
export ANTHROPIC_BASE_URL=https://api.taidatech.ai
export ANTHROPIC_AUTH_TOKEN=sk-your-nmd-api-key
# The client automatically appends the /v1/messages path.

Responses API clients

The platform provides POST /v1/responses and /v1/responses/compact for clients such as Codex CLI and the OpenAI Agents SDK.

# ~/.codex/config.toml
model = "auto"
model_provider = "taida"

[model_providers.taida]
name = "Taida Tech"
base_url = "https://api.taidatech.ai/v1"
env_key = "NEXUS_API_KEY"
wire_api = "responses"

Authentication & API Keys

Every request must include an API key. The key identifies the account and controls model permissions, rate limits, and available balance.

NEXUS_API_KEY=sk-your-nmd-api-key

Models & Pricing

Usage is billed in USD at the prices shown below. The models available to your account are returned by GET /v1/models.

Chat Completions API

Text models use the OpenAI-compatible Chat Completions endpoint. To switch models, change the model field in the request body.

curl https://api.taidatech.ai/v1/chat/completions \
  -H "Authorization: Bearer $NEXUS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen3.6-flash",
    "messages": [
      {"role": "system", "content": "You are a professional development assistant. "},
      {"role": "user", "content": "Explain in three sentences what is MaaS. "}
    ],
    "stream": false
  }'

Internet search

Some models support native web search. Add enable_search: true to request current information:

curl https://api.taidatech.ai/v1/chat/completions \
  -H "Authorization: Bearer $NEXUS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen3.7-plus",
    "messages": [{"role": "user", "content": "What is in the technology news today?"}],
    "enable_search": true
  }'

• Web search is supported by qwen3.6-flash, qwen3.6-plus, qwen3.7-plus, qwen3.7-max, qwen3.5-omni-flash, qwen3.5-omni-plus, deepseek-v4-flash, deepseek-v4-pro, kimi-k2.7-code, and auto.
• glm-5.2 does not support this parameter and returns invalid_parameter_error.
• The qwen3.5-omni series requires streaming mode (stream: true) for web search. The gateway injects the required upstream search settings automatically.
• In Model Studio, supported models use web search by default. You can disable it from the conversation toolbar.

Image understanding

curl https://api.taidatech.ai/v1/chat/completions \
  -H "Authorization: Bearer $NEXUS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen3.7-plus",
    "messages": [{
      "role": "user",
      "content": [
        {"type": "text", "text": "Describe the image and extract its text."},
        {"type": "image_url", "image_url": {"url": "https://example.com/image.jpg"}}
      ]
    }]
  }'

Smart Routing (auto)

Set model to auto and the gateway picks a suitable default model for the endpoint you call. Billing always follows the model that was actually used, and the usage ledger records it.

curl https://api.taidatech.ai/v1/chat/completions \
  -H "Authorization: Bearer $NEXUS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "auto",
    "messages": [{"role": "user", "content": "Write a short Python example and explain how it works"}]
  }'

• /v1/chat/completions, /v1/completions, and /v1/responses route to qwen3.6-flash.
• /v1/embeddings routes to qwen3.7-text-embedding.
• /v1/images/generations routes to qwen-image-2.0; /v1/videos routes to wan2.7-t2v.
• /v1/audio/speech routes to qwen3-tts-flash.
• Specify a fixed model ID when consistent model behavior is required.

Embeddings API

qwen3.7-text-embedding converts text into vector embeddings for knowledge-base retrieval, semantic search, and clustering. It is billed by tokens.

curl https://api.taidatech.ai/v1/embeddings \
  -H "Authorization: Bearer $NEXUS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen3.7-text-embedding",
    "input": ["First paragraph", "Second paragraph"]
  }'

• Embedding models are API-only and do not appear in the Playground conversation list.

Image Generation

Image models use the standard image-generation endpoint and are billed per image. Supported families include Qwen Image, Wan, and Z-Image.

curl https://api.taidatech.ai/v1/images/generations \
  -H "Authorization: Bearer $NEXUS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen-image-2.0",
    "prompt": "A clear glass cube in a white studio, product photography",
    "size": "1024x1024",
    "n": 1
  }'

• n sets the number of images. Cost is based on the number generated.
• Large images (such as 2048×2048) take approximately 45–70 seconds. Set the client timeout to at least 120 seconds.
• Generated results are subject to upstream content-safety review. Rejected requests may return HTTP 400 and are not charged.
• Download generated media promptly and store it in your own object storage when using the public API.

Video Generation

After creating a task, use the returned task ID to check its status. Video usage is billed by duration in seconds. HappyHorse supports text-to-video (t2v), image-to-video (i2v), and reference-to-video (r2v).

curl https://api.taidatech.ai/v1/videos \
  -H "Authorization: Bearer $NEXUS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "wan2.7-t2v",
    "prompt": "A spaceship sweeps past the clouds as the camera slowly moves forward, cinematic lighting",
    "size": "1280*720",
    "duration": 5
  }'

curl https://api.taidatech.ai/v1/videos/TASK_ID \
  -H "Authorization: Bearer $NEXUS_API_KEY"

Speech Synthesis API

qwen3-tts-flash converts text to natural speech. Usage is billed by character ($10 per million characters), and the current output format is WAV.

curl https://api.taidatech.ai/v1/audio/speech \
  -H "Authorization: Bearer $NEXUS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen3-tts-flash",
    "input": "Hello. Welcome to Taida Nexus speech synthesis.",
    "voice": "Cherry"
  }'

• Available voices: Cherry (default), Ethan, Nofish, Jennifer, Ryan, Katerina, Elias, Serena, Marcus, and Bella.
• voice selects the voice; speed adjusts speaking speed from 0.5 to 2.0 (default 1.0).
• Keep each request to 5,000 characters or fewer.

Common Errors

Open Source and Compliance

The platform gateway is based on the open-source New API project under AGPLv3. Taida customizations cover model adaptation and intelligent routing. In accordance with AGPLv3 Section 13, the complete modified gateway source is available here:

• Modified source repository: github.com/leonczh07-byte/new-api (taida branch, matching the online gateway).
• File-by-file modification list: TAIDA_MODIFICATIONS.md.
• The gateway remains subject to AGPLv3. All original copyright notices are retained.
• The platform console, billing, and account system are separate proprietary components that interact with the gateway through APIs.

For source-code compliance questions, open an issue in the repository above or email contact@taidatech.ai.