langchain-nativ 0.1.0

LangChain integration for Nativ — AI-powered localization tools

langchain-nativ

LangChain integration for Nativ — AI-powered localization.

Give any LangChain agent the ability to translate text, search translation memory, manage terminology, and more — all backed by your team's brand voice and style guides.

Installation

pip install langchain-nativ

Quick start

from langchain_nativ import NativTranslateTool

tool = NativTranslateTool()  # reads NATIV_API_KEY from env

result = tool.invoke({
    "text": "Hello world",
    "target_language": "French",
})
print(result)
# Bonjour le monde
# Rationale: Standard greeting translated with neutral register.

Use with a LangChain agent

from langchain_nativ import NativToolkit
from langchain_openai import ChatOpenAI
from langgraph.prebuilt import create_react_agent

tools = NativToolkit().get_tools()
llm = ChatOpenAI(model="gpt-4o")
agent = create_react_agent(llm, tools)

result = agent.invoke({
    "messages": [{
        "role": "user",
        "content": "Translate 'Welcome back!' to French, German, and Japanese",
    }]
})

Available tools

Tool	Description
NativTranslateTool	Translate text with cultural adaptation
NativTranslateBatchTool	Translate multiple texts to one language
NativSearchTranslationMemoryTool	Fuzzy-search existing translations
NativAddTranslationMemoryEntryTool	Store an approved translation for reuse
NativGetLanguagesTool	List configured target languages
NativGetStyleGuidesTool	Get style guide content
NativGetBrandVoiceTool	Get the brand voice prompt
NativGetTranslationMemoryStatsTool	Get TM statistics

Configuration

All tools accept api_key and base_url parameters:

tool = NativTranslateTool(api_key="nativ_...", base_url="https://api.usenativ.com")

Or set the environment variable:

export NATIV_API_KEY=nativ_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

Use individual tools

from langchain_nativ import NativSearchTranslationMemoryTool

search = NativSearchTranslationMemoryTool()
print(search.invoke({"query": "Welcome"}))
# Found 3 match(es):
# - [98% exact] "Welcome" → "Bienvenue"
# - [85% fuzzy] "Welcome back" → "Content de vous revoir"
# ...

License

MIT