Perplexity AI搜索API教程

想让自己的应用返回带引用的实时网络答案。

Perplexity AI Search API Guide 2026: Build a Research Assistant in 20 Minutes

The Perplexity API is the easiest way to add live web search to an app. Unlike a plain LLM API, it runs a real web search before answering and returns citations. By August 2026 the API supports the sonar and sonar-pro models, with sonar-deep-research in beta for multi-step research.

What You Get

• Real-time web answers with cited sources
• Token usage comparable to a normal LLM call
• A Python SDK and REST endpoint
• Usage-based pricing with a free tier for new accounts

Step 1: Get a Key

Sign up at perplexity.ai, go to Settings > API, and create a key. As of mid-2026, new accounts get $5 in free credits - roughly 2,000 sonar queries.

Step 2: The Minimal Call

from openai import OpenAI

client = OpenAI(api_key="pplx-...", base_url="https://api.perplexity.ai")

response = client.chat.completions.create(
    model="sonar",
    messages=[
        {"role": "system", "content": "Answer concisely with citations."},
        {"role": "user", "content": "What are the latest specs of the Raspberry Pi 5?"}
    ],
)
print(response.choices[0].message.content)

The response includes a citations field with URLs. That is the whole trick - your app now answers with live sources.

Step 3: Force Fresh Search

Set "search_recency_filter": "week" in the extra body to force a recent search window. Default is month.

Step 4: Handle Citations

for c in response.citations:
    print(c.url)

Store the URLs and render them as links - users trust answers they can verify.

Cost Reality Check (August 2026)

FAQ

Is the Perplexity API free? There is a $5 trial credit; after that it is usage-based.

Can I use it without the OpenAI SDK? Yes, the REST endpoint is https://api.perplexity.ai/chat/completions with the same body format.

How is it different from the ChatGPT API? Perplexity performs live web search before answering; a plain LLM API only uses its training data.