LangChain Nebius Integration

0.1.3 · active · verified Sun Apr 12

This package provides LangChain integration for Nebius AI Studio, enabling seamless use of Nebius AI Studio's chat and embedding models within LangChain. It offers classes for chat models (ChatNebius), embedding models (NebiusEmbeddings), and a retriever (NebiusRetriever), facilitating common LLM application patterns like RAG. The current version is 0.1.3.

Warnings

gotcha API Key Management: Nebius requires an API key for authentication. It's recommended to set it as an environment variable named `NEBIUS_API_KEY` rather than passing it directly in code for security reasons.
Fix: Set `export NEBIUS_API_KEY='your_api_key'` in your environment or use a `.env` file and `dotenv`.
gotcha LangChain Version Compatibility: As the LangChain ecosystem (especially `langchain-core`) evolves rapidly, ensure your installed `langchain-nebius` version is compatible with your core LangChain packages to avoid import errors or unexpected behavior.
Fix: Refer to the `langchain-nebius` GitHub repository or PyPI page for explicit `langchain-core` version requirements. Use `pip install -U langchain-nebius` to get the latest compatible version.
gotcha Model String Variations: The `model` parameter for `ChatNebius` and `NebiusEmbeddings` expects a specific string identifier. These identifiers may vary or new models may become available. Refer to the Nebius AI Studio documentation for the most current list of supported models.
Fix: Consult the official Nebius AI Studio documentation for up-to-date model names and availability. Avoid hardcoding model names if they are subject to frequent changes.
gotcha Rate Limiting and Context Length: Like most LLM providers, Nebius AI Studio may impose rate limits and have maximum context window sizes for its models. Hitting these limits can lead to errors.
Fix: Implement retry mechanisms with exponential backoff for API calls. Monitor token usage and truncate input or summarize history if close to context limits. Review Nebius AI Studio's API documentation for specific rate limits and model context window sizes.

Install

pip install langchain-nebius Install stable version

Imports

ChatNebius
```
from langchain_nebius import ChatNebius
```

NebiusEmbeddings

from langchain_nebius import NebiusEmbeddings

NebiusRetriever

from langchain_nebius import NebiusRetriever

NebiusRetrievalTool

from langchain_nebius import NebiusRetrievalTool

Quickstart

This example demonstrates how to initialize `ChatNebius` and invoke it with a simple query. Ensure your `NEBIUS_API_KEY` is set as an environment variable for secure credential management. You may need to select a specific model available in Nebius AI Studio.

import os
from langchain_nebius import ChatNebius
from langchain_core.messages import HumanMessage

# Set your Nebius API key as an environment variable
# os.environ["NEBIUS_API_KEY"] = "your_nebius_api_key"

chat = ChatNebius(
    api_key=os.environ.get('NEBIUS_API_KEY', ''),
    model="Qwen/Qwen3-14B", # Choose an available model from Nebius AI Studio
    temperature=0.6
)

response = chat.invoke([
    HumanMessage(content="What is 1 + 1?")
])

print(response.content)

view raw JSON →