langchain-talor-serp

LangChain integration for TalorData's SERP APIs

Installation • Quick Start • Tools • Resources

LangChain integration for the Talor SERP API.

This package provides:

• TalorSerpAPIWrapper for direct sync and async API access
• TalorSerpTool for creating LangChain tools
• bundled engine schemas for 30+ search engines
• support for search, history, and statistics endpoints

Overview

langchain-talor-serp provides LangChain tools for TalorData's SERP APIs, enabling your AI agents to:

• Search - Query search engines with geo-targeting and language customization
• Inspect engines - Discover supported engines and engine-specific parameters
• Query history - Fetch SERP request history with filters
• View statistics - Retrieve usage statistics by date range and engine

Installation

pip install langchain-talor-serp

Quick start

1. Get your API key

Sign up at TalorData and get your API key from the dashboard.

2. Set up authentication

import os
os.environ["TALOR_API_KEY"] = "your-token"

3. Modern agent usage

from langchain_talor_serp import TalorSerpTool
from langchain_openai import ChatOpenAI

llm = ChatOpenAI(model="gpt-4o-mini", temperature=0)
tool = TalorSerpTool.from_env()

# Tool calling without langchain_classic agents
model_with_tools = llm.bind_tools([tool])
response = model_with_tools.invoke("Search for the latest LangChain news")
print(response)

4. Search tool

from langchain_talor_serp import TalorSerpTool

search_tool = TalorSerpTool.from_env()

result = search_tool.invoke({
    "query": "LangChain tutorial",
    "engine": "google",
    "params": {
        "gl": "us",
        "hl": "en",
        "device": "desktop",
    },
})

print(result)

Search parameters:

• query: required search query text
• engine: optional engine key such as google, google_news, google_images, bing, duckduckgo
• params: optional engine-specific parameter object
• common params fields include gl, hl, device, location, and no_cache
• use talor_serp_list_engines to inspect detailed parameters for a specific engine

params also accepts a JSON string when returned by a model tool call, for example:

result = search_tool.invoke({
    "query": "LangChain tutorial",
    "engine": "google",
    "params": "{\"hl\": \"zh-CN\", \"gl\": \"cn\"}",
})

5. History tool

from langchain_talor_serp import TalorSerpTool

history_tool = TalorSerpTool.history_from_env()

result = history_tool.invoke({
    "page": 1,
    "page_size": 20,
    "search_query": "langchain",
    "search_engine": "google",
    "status": "success",
    "timezone": "Asia/Shanghai",
})

print(result)

History parameters:

• page: page number, default 1
• page_size: page size, default 20
• search_query: optional keyword filter
• search_engine: optional engine filter such as google or bing
• status: all, success, or error
• start_time: optional unix timestamp in seconds
• end_time: optional unix timestamp in seconds
• timezone: optional timezone header such as Asia/Shanghai or +08:00

6. Statistics tool

from langchain_talor_serp import TalorSerpTool

statistics_tool = TalorSerpTool.statistics_from_env()

result = statistics_tool.invoke({
    "start_date": "2026-06-01",
    "end_date": "2026-06-05",
    "engines": "google,bing",
    "timezone": "+08:00",
})

print(result)

Statistics parameters:

• start_date: required, format YYYY-MM-DD
• end_date: required, format YYYY-MM-DD
• engines: optional comma-separated engine keys such as google,bing
• timezone: optional timezone offset such as +08:00

7. Bind multiple tools

from langchain_talor_serp import TalorSerpTool
from langchain_openai import ChatOpenAI

llm = ChatOpenAI(model="gpt-4o-mini", temperature=0)
tools = TalorSerpTool.tools_from_env()

model_with_tools = llm.bind_tools(tools)
response = model_with_tools.invoke("Show my SERP usage statistics for 2026-06-01 to 2026-06-05")
print(response)

bind_tools() only lets the model generate tool_calls. To actually execute the selected tool, either:

• call tool.invoke(...) directly, or
• use an agent workflow that executes tools automatically

Tools

• talor_serp_search - search the web with engine-specific parameters
• talor_serp_list_engines - inspect supported engines and detailed parameter schemas
• talor_serp_history - query historical SERP requests
• talor_serp_statistics - query usage statistics for a date range

Compatibility note

If you are using modern package versions such as:

• langchain-core>=1.0
• langchain-classic>=1.0
• langchain-openai>=1.0

avoid langchain_classic.agents.create_openai_functions_agent() and other legacy initialize_agent / AgentType.OPENAI_FUNCTIONS flows with chat models. Those classic agent paths call llm.invoke(..., callbacks=...), while modern BaseChatModel.invoke() forwards callbacks through config, which can lead to:

TypeError: BaseLLM.generate_prompt() got multiple values for keyword argument 'callbacks'

Use one of these approaches instead:

• llm.bind_tools([tool]) for direct tool calling
• langchain.agents.create_agent(...) for new agent workflows
• agent.invoke(...) instead of deprecated agent.run(...)

Features

• compatible with LangChain tool workflows
• supports multiple Talor SERP engines such as Google, Bing, Yandex, and DuckDuckGo
• exposes engine schema metadata for parameter-aware integrations
• includes helper APIs for usage history and statistics

Resources

• PyPI: langchain-talor-serp
• TalorData: talordata.com