DeepSeek V4 Pro API with APIDot

DeepSeek V4 Pro API examples with APIDot: cURL, Node.js, request fields, responses, and production notes.

Try on APIDot | Get API Key | API Docs | Pricing | Main Examples

Why this repo exists

DeepSeek V4 Pro API for 1M-token long-context text chat, Pro-grade deep reasoning, complex coding, agentic planning, research synthesis, and OpenAI-compatible chat completions.

This repository turns that APIDot workflow into runnable server-side examples: a cURL request, a native Node.js example, request fields, response handling notes, and production integration guardrails.

Overview

DeepSeek V4 Pro is the high-depth Pro-tier model in the DeepSeek V4 text series. It is built for million-token reasoning, complex coding, full-codebase analysis, dense document review, and long-horizon planning where the task needs more depth than the Flash variant.

Choose DeepSeek V4 Pro for difficult reasoning, multi-file debugging, research synthesis, legal or financial review, and workflows where answer quality matters more than the lowest token cost. DeepSeek V4 Flash remains the better default for high-volume long-context traffic that does not need Pro-level depth.

On APIDot, use model ID deepseek-v4-pro through the existing OpenAI-compatible POST /v1/chat/completions path with Bearer authentication. The page supports streaming responses, usage fields for prompt and completion tokens, and separate APIDot input/output token pricing.

Endpoint

POST /v1/chat/completions

Send OpenAI-compatible chat completions requests to DeepSeek V4 Pro through APIDot.

Best practices

• Use deepseek-v4-pro for 1M-token long-context text chat, complex code review, research synthesis, and long-horizon agent planning.
• Keep reasoning-mode choices in your system prompt or user prompt. APIDot does not add a dedicated thinking_mode field for this model.
• Read assistant text from data.choices[0].message.content when the response is wrapped by APIDot, or choices[0].message.content for raw OpenAI-compatible responses.
• Read input and output usage from usage.prompt_tokens and usage.completion_tokens so billing can be checked separately.
• Set stream: true only when your client is prepared to consume server-sent chat completion chunks.

Quick start

cp .env.example .env
# Edit .env and set APIDOT_API_KEY
cd node
npm start

Use curl/generate.md when you want a copy-paste cURL request instead of Node.js.

Minimal API request

Submit to APIDot:

curl --fail-with-body --request POST \
  --url https://api.apidot.ai/v1/chat/completions \
  --header "Authorization: Bearer $APIDOT_API_KEY" \
  --header "Content-Type: application/json" \
  --data '{
  "model": "deepseek-v4-pro",
  "messages": [
    {
      "role": "system",
      "content": "You are a concise technical assistant."
    },
    {
      "role": "user",
      "content": "Summarize the migration risks in five bullets."
    }
  ],
  "max_tokens": 512
}'

Primary payload:

{
  "model": "deepseek-v4-pro",
  "messages": [
    {
      "role": "system",
      "content": "You are a concise technical assistant."
    },
    {
      "role": "user",
      "content": "Summarize the migration risks in five bullets."
    }
  ],
  "max_tokens": 512
}

Request variants

deepseek-v4-pro

{
  "model": "deepseek-v4-pro",
  "messages": [
    {
      "role": "system",
      "content": "You are a concise technical assistant."
    },
    {
      "role": "user",
      "content": "Summarize the migration risks in five bullets."
    }
  ],
  "max_tokens": 512
}

Request parameters

Field	Type	Required	Description
model	string	yes	Target model ID. Use deepseek-v4-pro.
messages	object[]	yes	Conversation array in OpenAI-compatible chat format.
messages[].role	string	yes	Message role. Use system, user, or assistant.
messages[].content	string	yes	Text content for the message.
temperature	number	no	Sampling temperature from 0 to 2 when you choose to override defaults.
top_p	number	no	Nucleus sampling value from 0 to 1.
max_tokens	integer	no	Maximum output tokens to return.
stream	boolean	no	Set true to receive streaming chat completion chunks.

Response fields

Field	Type	Description
code	integer	APIDot status code when the response is wrapped. Successful calls return 200.
data.id	string	Chat completion response ID.
data.choices[].message.content	string	Generated assistant text.
data.usage.prompt_tokens	integer	Input token count billed at the DeepSeek V4 Pro input rate.
data.usage.completion_tokens	integer	Output token count billed at the DeepSeek V4 Pro output rate.
data.usage.total_tokens	integer	Total token count across input and output.

Example response

{
  "code": 200,
  "data": {
    "id": "chatcmpl_1234567890",
    "object": "chat.completion",
    "model": "deepseek-v4-pro",
    "choices": [{"index": 0, "message": {"role": "assistant", "content": "The key risks are..."}, "finish_reason": "stop"}],
    "usage": {"prompt_tokens": 64, "completion_tokens": 96, "total_tokens": 160}
  }
}

Error classes

Code	Name	Description
400	invalid_request_error	The request body is missing required chat fields or includes invalid values.
401	authentication_error	Missing, expired, or invalid Bearer API key.
402	insufficient_credits	The account does not have enough credits for the requested token usage.
404	model_not_found	The selected public model ID is not mapped in the backend vendor feature configuration.
500	service_error	The model request failed or timed out.

Production notes

• Keep APIDot API keys in server-side environment variables.
• Persist selected model, request payload, user ID, and response metadata together for support and cost review.
• Avoid logging API keys, private prompts, customer data, private media URLs, or internal callback URLs.
• Validate request fields before sending traffic to APIDot.
• Retry transient network failures with backoff, but do not retry unchanged invalid payloads.

Common mistakes

• Committing a real API key or .env file.
• Calling APIDot directly from browser code instead of a server-side environment.
• Logging private prompts, customer data, API keys, or internal callback URLs.
• Reusing request fields from another model without checking this model's docs.

FAQ

Which DeepSeek V4 Pro model ID can I send?

Send deepseek-v4-pro in the model field.

What authentication header should I use?

Use Authorization: Bearer $APIDOT_API_KEY when calling /v1/chat/completions.

Is this endpoint OpenAI-compatible?

Yes. It accepts the standard chat completions payload shape for text chat.

Can I stream responses?

Yes. Set stream: true to receive streaming chat completion chunks.

Related links

• Website: https://apidot.ai
• Docs: https://apidot.ai/docs
• DeepSeek V4 Pro docs: https://apidot.ai/docs/deepseek-v4-pro
• DeepSeek V4 Pro model page: https://apidot.ai/models/deepseek-v4-pro
• GitHub organization: https://github.com/APIDotAI
• Main examples: https://github.com/APIDotAI/apidot-examples