Context Grounding¶

Context Grounding Service allows you to:

Search through indexed documents using natural language queries
Ground LLM responses in your organization's specific information
Retrieve context-relevant documents for various applications

You will need to create an index in Context Grounding to use this feature. To create an index go to organization Admin -> AI Trust Layer -> Context Grounding. There you can create a new index and add documents to it. See the full documentation here for more details.

ContextGroundingRetriever¶

The ContextGroundingRetriever is a document retrieval system that uses vector search to efficiently find and retrieve relevant information from your document store.

Basic Usage¶

Create a simple retriever by specifying an index name:

from uipath_langchain.retrievers import ContextGroundingRetriever

retriever = ContextGroundingRetriever(index_name = "Company Policy Context")
print(retriever.invoke("What is the company policy on remote work?"))

Integration with LangChain Tools¶

You can easily integrate the retriever with LangChain's tool system:

from langchain.agents import create_react_agent
from langchain.tools.retriever import create_retriever_tool
from uipath_langchain.retrievers import ContextGroundingRetriever

retriever = ContextGroundingRetriever(index_name = "Company Policy Context")
retriever_tool = create_retriever_tool(
    retriever,
    "ContextforInvoiceDisputeInvestigation",
   """
   Use this tool to search the company internal documents for information about policies around dispute resolution.
   Use a meaningful query to load relevant information from the documents. Save the citation for later use.
   """
)

# You can use the tool in your agents
model = OpenAI()
tools = [retriever_tool]
agent = create_react_agent(model, tools, prompt="Answer user questions as best as you can using the search tool.")

Advanced Usage¶

For complex applications, the retriever can be combined with other LangChain components to create robust document QA systems, agents, or knowledge bases.

ContextGroundingVectorStore¶

ContextGroundingVectorStore is a vector store implementation designed for context-aware document retrieval. It allows you to perform semantic searches and create retrieval chains with language models.

Searching Documents¶

The vector store supports various search methods:

from uipath_langchain.vectorstores.context_grounding_vectorstore import ContextGroundingVectorStore

vectorstore = ContextGroundingVectorStore(index_name="Company policy")

# Perform semantic searches with distance scores
docs_with_scores = vectorstore.asimilarity_search_with_score(query="What is the company policy on data storage?", k=5)

# Perform a similarity search with relevance scores
docs_with_relevance_scores = await vectorstore.asimilarity_search_with_relevance_scores(query=query, k=5)

Creating a Retrieval Chain¶

You can integrate the vector store into a retrieval chain with a language model:

# Run a retrieval chain
model = UiPathAzureChatOpenAI(model="gpt-4o-2024-08-06", max_retries=3)
retrieval_chain = create_retrieval_chain(vectorstore=vectorstore, model=model)

query = "What is the ECCN for a laptop?"
result = retrieval_chain(query)