langchain-scavio 2.2

LangChain integration for Scavio Search API -- real-time web search with knowledge graphs, multi-platform support

langchain-scavio

LangChain integration for the Scavio Search API. Real-time structured data from Google, Amazon, Walmart, and YouTube — all through a single package.

Why Scavio? Multi-platform coverage, structured knowledge graph data, and competitive pricing at $0.005/credit.

Installation

pip install langchain-scavio

Tools

Tool	Description
ScavioSearch	Google web search with knowledge graphs, PAA questions, news
ScavioAmazonSearch	Search Amazon product listings
ScavioAmazonProduct	Fetch full details for an Amazon product by ASIN
ScavioWalmartSearch	Search Walmart product listings
ScavioWalmartProduct	Fetch full details for a Walmart product by ID
ScavioYouTubeSearch	Search YouTube videos with duration/date/type filters
ScavioYouTubeMetadata	Fetch metadata for a YouTube video by video ID
ScavioYouTubeTranscript	Fetch the transcript of a YouTube video

Quick Start

Get your API key at dashboard.scavio.dev.

import os
from langchain_scavio import ScavioSearch

os.environ["SCAVIO_API_KEY"] = "sk_live_..."

tool = ScavioSearch()
result = tool.invoke({"query": "best python web frameworks 2026"})

Use with LangGraph

from langgraph.prebuilt import create_react_agent
from langchain_openai import ChatOpenAI
from langchain_scavio import (
    ScavioSearch,
    ScavioAmazonSearch, ScavioAmazonProduct,
    ScavioWalmartSearch,
    ScavioYouTubeSearch, ScavioYouTubeTranscript,
)

agent = create_react_agent(
    ChatOpenAI(model="gpt-4o"),
    tools=[
        ScavioSearch(max_results=5),
        ScavioAmazonSearch(max_results=5),
        ScavioAmazonProduct(),
        ScavioWalmartSearch(max_results=5),
        ScavioYouTubeSearch(max_results=5),
        ScavioYouTubeTranscript(),
    ],
)

response = agent.invoke({
    "messages": [{"role": "user", "content": "Find me a Python book on Amazon under $30"}]
})

Async Support

All tools support async invocation:

result = await tool.ainvoke({"query": "async python frameworks"})

Configuration

Google Search

from langchain_scavio import ScavioSearch

tool = ScavioSearch(
    scavio_api_key="sk_live_...",       # or SCAVIO_API_KEY env var
    max_results=5,
    light_request=None,                  # None=light/1 credit, False=full/2 credits
    include_knowledge_graph=True,
    include_questions=True,
    include_related=False,
    country_code="us",
    language="en",
    search_type="classic",               # classic|news|maps|images|lens
    device="desktop",
)

Amazon

from langchain_scavio import ScavioAmazonSearch, ScavioAmazonProduct

search = ScavioAmazonSearch(
    max_results=5,
    pages=1,                             # number of result pages to fetch
    domain="com",                        # com|co.uk|de|fr|co.jp|ca|...
)

product = ScavioAmazonProduct()
result = product.invoke({"query": "B08N5WRWNW"})  # query = ASIN

Walmart

from langchain_scavio import ScavioWalmartSearch, ScavioWalmartProduct

search = ScavioWalmartSearch(max_results=5)
result = search.invoke({
    "query": "air fryer",
    "sort_by": "price_low",              # best_match|price_low|price_high|best_seller
    "max_price": 5000,                   # in cents
    "fulfillment_speed": "2_days",       # today|tomorrow|2_days|anytime
})

product = ScavioWalmartProduct()
result = product.invoke({"product_id": "123456789"})

YouTube

from langchain_scavio import ScavioYouTubeSearch, ScavioYouTubeMetadata, ScavioYouTubeTranscript

search = ScavioYouTubeSearch(max_results=5)
result = search.invoke({
    "query": "python tutorial",
    "duration": "medium",                # short|medium|long
    "upload_date": "this_month",         # last_hour|today|this_week|this_month|this_year
    "sort_by": "view_count",             # relevance|date|view_count|rating
    "video_type": "video",               # video|channel|playlist
})

metadata = ScavioYouTubeMetadata()
result = metadata.invoke({"video_id": "dQw4w9WgXcQ"})

transcript = ScavioYouTubeTranscript(max_segments=200)
result = transcript.invoke({"video_id": "dQw4w9WgXcQ", "language": "en"})

Agent-Controllable Parameters

ScavioSearch

Parameter	Type	Description
query	str	Search query
search_type	classic|news|maps|images|lens	Type of search
country_code	str	ISO 3166-1 alpha-2
language	str	ISO 639-1
device	desktop|mobile	Device type
page	int	Result page number

ScavioAmazonSearch

Parameter	Type	Description
query	str	Product search query
domain	str	Amazon domain (com, co.uk, de, ...)
sort_by	str	most_recent|price_low_to_high|price_high_to_low|featured|average_review|bestsellers
start_page	int	Page number
category_id	str	Category filter
country / language / currency	str	Localization
zip_code	str	Local pricing

ScavioWalmartSearch

Parameter	Type	Description
query	str	Product search query
sort_by	str	best_match|price_low|price_high|best_seller
min_price / max_price	int	Price range in cents
fulfillment_speed	str	today|tomorrow|2_days|anytime
delivery_zip	str	Delivery ZIP code

ScavioYouTubeSearch

Parameter	Type	Description
query	str	Search query
upload_date	str	last_hour|today|this_week|this_month|this_year
video_type	str	video|channel|playlist
duration	str	short|medium|long
sort_by	str	relevance|date|view_count|rating
hd / subtitles / live	bool	Content filters

Error Handling

• Empty results raise ToolException with actionable suggestions for the LLM
• API errors return {"error": "message"} without crashing the agent
• handle_tool_error=True ensures LangChain passes errors to the LLM as context

Architecture

ScavioBaseAPIWrapper                      # Auth, headers, sync/async HTTP POST
  +-- ScavioSearchAPIWrapper              # -> /api/v1/google
  +-- ScavioAmazonSearchAPIWrapper        # -> /api/v1/amazon/search
  +-- ScavioAmazonProductAPIWrapper       # -> /api/v1/amazon/product
  +-- ScavioWalmartSearchAPIWrapper       # -> /api/v1/walmart/search
  +-- ScavioWalmartProductAPIWrapper      # -> /api/v1/walmart/product
  +-- ScavioYouTubeSearchAPIWrapper       # -> /api/v1/youtube/search
  +-- ScavioYouTubeMetadataAPIWrapper     # -> /api/v1/youtube/metadata
  +-- ScavioYouTubeTranscriptAPIWrapper   # -> /api/v1/youtube/transcript

Each tool splits parameters into init-only (developer-controlled, e.g. max_results, domain) and LLM-controllable (passed via args_schema at invocation time, e.g. query, sort_by).

Migrating from Tavily

- from langchain_tavily import TavilySearch
+ from langchain_scavio import ScavioSearch

- tool = TavilySearch(max_results=5)
+ tool = ScavioSearch(max_results=5)

See the full migration guide for parameter mapping and feature comparison.

License

MIT