DeepSeek API Integration

DeepSeek API Integration

DeepSeek is a Chinese LLM provider with models competing with GPT-4o at significantly lower cost. DeepSeek-R1 is a reasoning model with open weights comparable to OpenAI's o1. DeepSeek Coder V2 is a specialized code model. Important: data is processed in China, which is critical for compliance-sensitive tasks.

Basic Integration (OpenAI-Compatible API)

from openai import OpenAI

# DeepSeek is fully compatible with OpenAI SDK
client = OpenAI(
    api_key="DEEPSEEK_API_KEY",
    base_url="https://api.deepseek.com",
)

# Chat
response = client.chat.completions.create(
    model="deepseek-chat",  # DeepSeek-V3
    messages=[
        {"role": "system", "content": "You are an experienced Python developer"},
        {"role": "user", "content": "Write an async function for batch API requests"},
    ],
    temperature=0.1,
)
print(response.choices[0].message.content)

# Reasoning (deepseek-reasoner = R1)
response = client.chat.completions.create(
    model="deepseek-reasoner",
    messages=[{"role": "user", "content": "Prove that sqrt(2) is irrational"}],
)
# R1 returns reasoning_content (chain of thought) + content (answer)
print(response.choices[0].message.reasoning_content)  # Reasoning
print(response.choices[0].message.content)             # Final answer

Streaming

with client.chat.completions.stream(
    model="deepseek-chat",
    messages=[{"role": "user", "content": "Long answer..."}],
) as stream:
    for chunk in stream.text_stream:
        print(chunk, end="", flush=True)

Code Completion (FIM) — DeepSeek Coder

response = client.completions.create(
    model="deepseek-chat",
    prompt="<｜fim▁begin｜>def calculate_tax(income: float",
    suffix="<｜fim▁end｜>",
    max_tokens=128,
    stop=["<｜fim▁end｜>"],
)
print(response.choices[0].text)

Cost of DeepSeek (2025)

For comparison: GPT-4o $2.50/$10, making DeepSeek 5–10× cheaper with comparable quality.

When to Use

• Tasks requiring computation or deep analysis — R1
• Code, SQL, data analysis — DeepSeek-V3 or Coder V2
• High-load scenarios with price sensitivity — DeepSeek-V3
• Compliance requirements with data storage in RF/EU — not suitable

Local Deployment

ollama pull deepseek-r1:7b   # 4.7 GB
ollama pull deepseek-r1:70b  # 43 GB (requires GPU with 40GB VRAM)

Timeline

• Basic integration: 0.5 day (OpenAI-compatible API)
• Quality testing on specific tasks: 1–2 days