Search Tools (SearxNG)¶

The SearxngTools toolkit provides web search via a SearxNG instance and returns results as JSON strings.

It supports:

Multiple categories (general, news, images, videos, …)
Optional limited content fetching (disabled by default)
Optional file downloading/processing for a small allowlist (PDF/TXT/MD)

🤖 AI Agent Setup (Agno)¶

from agno.agent import Agent
from enhancedtoolkits import SearxngTools

agent = Agent(
    name="Research Assistant",
    model="gpt-4",
    tools=[
        SearxngTools(
            host="http://searxng:8080",
            max_results=10,
            enable_content_fetching=False,  # default
        )
    ],
)

⚙️ Configuration¶

Constructor parameters for SearxngTools:

Parameter	Type	Default	Notes
`host`	`str`	required	SearxNG base URL (e.g. `http://searxng:8080`)
`max_results`	`int \\| None`	`20`	Clamped internally (1..30)
`timeout`	`int`	`30`	Clamped internally (5..120)
`enable_content_fetching`	`bool \\| None`	`False`	If `True`, fetches at most 3 result pages (general/news/files only)
`enable_file_downloads`	`bool \\| None`	`True`	Enables file handling for supported types
`max_file_size_mb`	`int`	`10`	Clamped internally (1..500)
`file_download_timeout`	`int`	`60`	Clamped internally (10..300)
`byparr_enabled`	`bool \\| None`	`True`	Used only for content fetching; can be auto-detected if set to `None`

Supported categories¶

See SearxngTools.SUPPORTED_CATEGORIES.

🔍 Available Functions¶

All functions return a JSON string encoding a list of result objects.

perform_web_search(query, max_results=None)
perform_news_search(query, max_results=None)
perform_image_search(query, max_results=None)
perform_video_search(query, max_results=None)
perform_category_search(query, category, max_results=None)

✅ Examples¶

Web search¶

from enhancedtoolkits import SearxngTools

search = SearxngTools(host="http://searxng:8080", max_results=5)
results_json = search.perform_web_search("agno ai agents")

Category search¶

results_json = search.perform_category_search(
    query="latest LLM safety papers",
    category="science",
    max_results=5,
)

Enable limited content fetching (first 3 results only)¶

search = SearxngTools(
    host="http://searxng:8080",
    max_results=10,
    enable_content_fetching=True,
)

results_json = search.perform_news_search("EU AI Act updates", max_results=5)

📝 Notes / Best Practices¶

Keep max_results small (5–10) unless you really need more.
Content fetching can produce large outputs; it is disabled by default and limited to the first 3 results.

API Reference¶

docs/api/searxng.md