# Building a RAG Pipeline

> How to use HushSearch to power a retrieval-augmented generation system.

Source: http://localhost:3000/docs/guides/rag-pipeline

A complete example: search → extract → embed → query.

## Step 1: Search for Relevant Documents

```python

def search_web(query: str, n: int = 5) -> list[dict]:
    response = requests.post(
        "https://api.hushsearch.io/v1/search/",
        headers={"X-API-Key": "hs_live_..."},
        json={"query": query, "num_results": n}
    )
    return response.json()["results"]
```

## Step 2: Extract Full Content

```python
def extract_pages(urls: list[str]) -> list[str]:
    docs = []
    for url in urls:
        r = requests.post(
            "https://api.hushsearch.io/v1/extract/",
            headers={"X-API-Key": "hs_live_..."},
            json={"url": url}
        )
        if r.ok:
            docs.append(r.json()["markdown_content"])
    return docs
```

## Step 3: Build Context and Query LLM

```python

def rag_answer(user_query: str) -> str:
    results  = search_web(user_query)
    urls     = [r["url"] for r in results[:3]]
    docs     = extract_pages(urls)
    context  = "\n\n---\n\n".join(docs[:3])

    client   = anthropic.Anthropic()
    message  = client.messages.create(
        model="claude-sonnet-4-6",
        max_tokens=1024,
        messages=[{
            "role": "user",
            "content": f"Context:\n{context}\n\nQuestion: {user_query}"
        }]
    )
    return message.content[0].text

print(rag_answer("What are the latest developments in AI agents?"))
```

## Credits Used

For a typical RAG query:
- 1 search call = **1 credit**
- 3 extract calls = **6 credits**
- **Total: 7 credits per query**

Or use `/v1/search-and-summarize/` to do it in one call for **5 credits**.
