Use Siftdog with LangChain and the Tavily SDK

In short: add api_base_url to the client. Everything else stays the same.

tavily-python

from tavily import TavilyClient

client = TavilyClient(api_key="your-siftdog-key", api_base_url="http://localhost:8000")

client.search("what is BM25", search_depth="advanced", max_results=5)
client.extract(urls=["https://en.wikipedia.org/wiki/Okapi_BM25"])
client.crawl("https://docs.python.org/3/library/asyncio.html", max_depth=1, limit=10)
client.map("https://docs.python.org/3/library/asyncio.html", max_depth=1)

The async client works the same way:

from tavily import AsyncTavilyClient

client = AsyncTavilyClient(api_key="your-siftdog-key", api_base_url="http://localhost:8000")
results = await client.search("python asyncio")

LangChain

With langchain-tavily:

from langchain_tavily import TavilyExtract, TavilySearch

search = TavilySearch(
    max_results=5, tavily_api_key="your-siftdog-key", api_base_url="http://localhost:8000"
)
search.invoke({"query": "what is BM25 ranking"})

extract = TavilyExtract(tavily_api_key="your-siftdog-key", api_base_url="http://localhost:8000")
extract.invoke({"urls": ["https://example.com"]})

These are regular LangChain tools, so they plug into agents like the originals.

Notes

• The key is any string when the server’s API_KEYS is empty, or one of those keys when it’s set.
• Scores come from BM25 blended with the search engines’ order, so they differ from Tavily’s.
• Not implemented: Tavily’s /research endpoint.

New to Siftdog? Start with how to self-host a Tavily alternative.