This will help you get started with Contextual AI’s Grounded Language Model chat models. To learn more about Contextual AI, please visit our documentation. This integration requires the contextual-client Python SDK. Learn more about it here.

​
Overview

This integration invokes Contextual AI’s Grounded Language Model.

​
Integration details

ClassPackageLocalSerializableJS supportPackage downloadsPackage latest
ChatContextuallangchain-contextual❌beta❌PyPI - DownloadsPyPI - Version

​
Model features

Tool callingStructured outputJSON modeImage inputAudio inputVideo inputToken-level streamingNative asyncToken usageLogprobs
❌❌❌❌❌❌❌❌❌❌

​
Setup

To access Contextual models you’ll need to create a Contextual AI account, get an API key, and install the langchain-contextual integration package.

​
Credentials

Head to app.contextual.ai to sign up to Contextual and generate an API key. Once you’ve done this set the CONTEXTUAL_AI_API_KEY environment variable: 
import getpass
import os

if not os.getenv("CONTEXTUAL_AI_API_KEY"):
    os.environ["CONTEXTUAL_AI_API_KEY"] = getpass.getpass(
        "Enter your Contextual API key: "
    )
If you want to get automated tracing of your model calls you can also set your LangSmith API key by uncommenting below: 
# os.environ["LANGSMITH_TRACING"] = "true"
# os.environ["LANGSMITH_API_KEY"] = getpass.getpass("Enter your LangSmith API key: ")

​
Installation

The LangChain Contextual integration lives in the langchain-contextual package: 
%pip install -qU langchain-contextual

​
Instantiation

Now we can instantiate our model object and generate chat completions. The chat client can be instantiated with these following additional settings:
ParameterTypeDescriptionDefault
temperatureOptional[float]The sampling temperature, which affects the randomness in the response. Note that higher temperature values can reduce groundedness.0
top_pOptional[float]A parameter for nucleus sampling, an alternative to temperature which also affects the randomness of the response. Note that higher top_p values can reduce groundedness.0.9
max_new_tokensOptional[int]The maximum number of tokens that the model can generate in the response. Minimum is 1 and maximum is 2048.1024
from langchain_contextual import ChatContextual

llm = ChatContextual(
    model="v1",  # defaults to `v1`
    api_key="",
    temperature=0,  # defaults to 0
    top_p=0.9,  # defaults to 0.9
    max_new_tokens=1024,  # defaults to 1024
)

​
Invocation

The Contextual Grounded Language Model accepts additional kwargs when calling the ChatContextual.invoke method. These additional inputs are:
ParameterTypeDescription
knowledgelist[str]Required: A list of strings of knowledge sources the grounded language model can use when generating a response.
system_promptOptional[str]Optional: Instructions the model should follow when generating responses. Note that we do not guarantee that the model follows these instructions exactly.
avoid_commentaryOptional[bool]Optional (Defaults to False): Flag to indicate whether the model should avoid providing additional commentary in responses. Commentary is conversational in nature and does not contain verifiable claims; therefore, commentary is not strictly grounded in available context. However, commentary may provide useful context which improves the helpfulness of responses.
# include a system prompt (optional)
system_prompt = "You are a helpful assistant that uses all of the provided knowledge to answer the user's query to the best of your ability."

# provide your own knowledge from your knowledge-base here in an array of string
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.",
]

# create your message
messages = [
    ("human", "What type of cats are there in the world and what are the types?"),
]

# invoke the GLM by providing the knowledge strings, optional system prompt
# if you want to turn off the GLM's commentary, pass True to the `avoid_commentary` argument
ai_msg = llm.invoke(
    messages, knowledge=knowledge, system_prompt=system_prompt, avoid_commentary=True
)

print(ai_msg.content)

​
Chaining

We can chain the Contextual Model with output parsers. 
from langchain_core.output_parsers import StrOutputParser

chain = llm | StrOutputParser

chain.invoke(
    messages, knowledge=knowledge, systemp_prompt=system_prompt, avoid_commentary=True
)

​
API reference

For detailed documentation of all ChatContextual features and configurations head to the Github page: https://github.com/ContextualAI//langchain-contextual