Skip to main content
The search_depth parameter controls the tradeoff between latency, cost, and content richness. Pick the right depth for each query instead of over-fetching.

​
Overview

DepthLatencyContentCreditsBest For
liteLowestTitles, URLs, snippets1 per searchURL discovery, filtering, high-volume pipelines
fastLowRich content2 per searchRAG, chatbots, AI agent loops
deepHigherFull real-time page extraction1 + 1 per pageResearch, due diligence, knowledge bases
fast mode is an enterprise feature. Contact sales to enable it for your account.

​
When to use each depth

​
Lite — scan and filter

Use lite when you need to scan many results quickly without reading full content. Returns titles, URLs, and short snippets. Good for:
  • URL discovery before targeted extraction
  • Filtering results by title or domain before deeper processing
  • High-volume monitoring where you only need to detect new mentions
  • Building URL lists for downstream tools like Extract or Crawl
{
  "query": "competitor product launches 2026",
  "search_depth": "lite",
  "max_results": 20
}

​
Fast — rich content for AI agents

Use fast when you need enough content to answer questions or feed an LLM, without the latency of scraping every page in real time. Returns rich content optimized for AI consumption. Good for:
  • RAG pipelines and chatbot grounding
  • Real-time agent workflows where latency matters
  • Q&A systems that need context beyond snippets
  • Any AI application that needs content, not just links
{
  "query": "python async best practices",
  "search_depth": "fast",
  "max_results": 5
}
fast mode requires an enterprise account and only works with focus: "general". Contact sales to get access.

​
Deep — full page extraction

Use deep when you need complete source material from every result. Each page is scraped in real time and returned as full content. Good for:
  • Research and due diligence requiring complete source text
  • Building comprehensive knowledge bases
  • Content analysis where snippets are insufficient
  • Legal or compliance workflows needing full page archives
{
  "query": "GDPR compliance requirements for SaaS",
  "search_depth": "deep",
  "max_results": 5,
  "output_format": "markdown"
}

​
Cost optimization

​
Start lite, go deeper when needed

The most cost-effective pattern: search with lite first, then use Extract on the specific URLs that matter. 
from nimble_python import Nimble

nimble = Nimble(api_key="YOUR-API-KEY")

# Step 1: Lite search to find relevant URLs
results = nimble.search(
    query="AI agent frameworks comparison",
    search_depth="lite",
    max_results=20
)

# Step 2: Extract full content from the top results only
for item in results.results[:3]:
    page = nimble.extract(
        url=item.url,
        output_format="markdown"
    )
    print(page.content)

​
Choose depth by query type

Not every query needs the same depth. Match depth to intent:
Query intentRecommended depthWhy
”Find URLs about X”liteOnly need links, not content
”What is X?”fastRich content is enough for a summary
”Summarize everything about X”deepNeed full page text for comprehensive analysis
”Monitor mentions of X”liteHigh-volume, only need detection
”Research X for a report”deepNeed complete source material

​
Combining depth with other features

​
Depth + domain filtering

Narrow your search to trusted sources before extracting content: 
{
  "query": "kubernetes best practices",
  "search_depth": "deep",
  "max_results": 5,
  "include_domains": ["kubernetes.io", "cloud.google.com", "docs.aws.amazon.com"]
}

​
Depth + time filtering

Combine depth with recency filters for targeted research: 
{
  "query": "AI regulation updates",
  "search_depth": "lite",
  "max_results": 20,
  "time_range": "week"
}

​
Depth + LLM answer

Add include_answer: true to any depth for an AI-generated summary with citations: 
{
  "query": "benefits of TypeScript over JavaScript",
  "search_depth": "lite",
  "max_results": 10,
  "include_answer": true
}

​
Next steps

Search

Full Search documentation with all features and topic modes

API Reference

Complete parameter documentation and response schemas