langchain-contextual 0.4.5

An integration package connecting Contextual and LangChain

langchain-contextual

This package contains the LangChain integration with Contextual AI.

Contextual AI provides state-of-the-art RAG components designed specifically for accurate and reliable enterprise AI applications. Founded by the inventors of RAG technology, our specialized components help innovative teams accelerate the development of production-ready AI applications that deliver responses with exceptional accuracy, even when processing millions of pages of documents.

This integration allows you to easily incorporate Contextual AI's Grounded Language Model and Instruction-Following Reranker into your LangChain workflows.

Features

This package provides access to two key components from Contextual AI:

• Grounded Language Model (GLM): The world's most grounded language model, engineered to minimize hallucinations by prioritizing faithfulness to retrieved knowledge. GLM delivers exceptional factual accuracy with inline attributions, making it ideal for enterprise RAG applications where reliability is critical.

• Instruction-Following Reranker: The first reranker that follows custom instructions to intelligently prioritize documents based on specific criteria like recency, source, or document type. Our reranker resolves conflicting information challenges in enterprise knowledge bases and outperforms competitors on industry benchmarks.

Installation

pip install -U langchain-contextual

And you should configure credentials by setting the following environment variables:

CONTEXTUAL_AI_API_KEY to your API key for Contextual AI

Using the Chat Models

The ChatContextual class exposes chat models like the Grounded Language Model (GLM) from Contextual.

from langchain_contextual import ChatContextual

llm = ChatContextual(
    model="v1",
    max_new_tokens=1024,
    temperature=0,
    top_p=0.9,
)

# only "human" and "ai" are accepted types of messages
# message types must alternative between "human" and "ai" if more than one message
messages = [
    ("human", "What type of cats are there in the world and what are the types?"),
]

knowledge = [
    "There are 2 types of dogs in the world: good dogs and best dogs.",
    "There are 2 types of cats in the world: good cats and best cats.",
]

llm.invoke(messages, knowledge=knowledge)

Using the Reranker

The ContextualRerank class exposes the reranker model from Contextual.

Example Usage

from langchain_core.documents import Document
from langchain_contextual import ContextualRerank

model = "ctxl-rerank-en-v1-instruct"

compressor = ContextualRerank(model=model)

query = "What is the current enterprise pricing for the RTX 5090 GPU for bulk orders?"

instruction = "Prioritize internal sales documents over market analysis reports. More recent documents should be weighted higher. Enterprise portal content supersedes distributor communications."

document_contents = [
    "Following detailed cost analysis and market research, we have implemented the following changes: AI training clusters will see a 15% uplift in raw compute performance, enterprise support packages are being restructured, and bulk procurement programs (100+ units) for the RTX 5090 Enterprise series will operate on a $2,899 baseline.",
    "Enterprise pricing for the RTX 5090 GPU bulk orders (100+ units) is currently set at $3,100-$3,300 per unit. This pricing for RTX 5090 enterprise bulk orders has been confirmed across all major distribution channels.",
    "RTX 5090 Enterprise GPU requires 450W TDP and 20% cooling overhead."
]

metadata = [
    {
        "Date": "January 15, 2025",
        "Source": "NVIDIA Enterprise Sales Portal",
        "Classification": "Internal Use Only"
    },
    {
        "Date": "11/30/2023",
        "Source": "TechAnalytics Research Group"
    },
    {
        "Date": "January 25, 2025",
        "Source": "NVIDIA Enterprise Sales Portal",
        "Classification": "Internal Use Only"
    }
]

documents = [
    Document(page_content=content, metadata=metadata[i])
    for i, content in enumerate(document_contents)
]
reranked_documents = compressor.compress_documents(
    query=query,
    instruction=instruction,
    documents=documents,
)