UiPath LLMs and Embeddings¶

This guide covers the UiPath-integrated Large Language Models (LLMs) and embedding models available in the UiPath LlamaIndex SDK.

Overview¶

The UiPath LlamaIndex SDK provides pre-configured LLM and embedding classes that integrate seamlessly with UiPath. These classes handle authentication, routing, and configuration automatically, allowing you to focus on building your agents.

Prerequisites¶

Before using these classes, ensure you have:

Authenticated with UiPath using uipath auth
Set up your environment variables (automatically configured after authentication)

UiPathOpenAI¶

The UiPathOpenAI class is a pre-configured Azure OpenAI client that routes requests through UiPath.

Available Models¶

The following OpenAI models are available through the OpenAIModel enum:

GPT_4_1_2025_04_14
GPT_4_1_MINI_2025_04_14
GPT_4_1_NANO_2025_04_14
GPT_4O_2024_05_13
GPT_4O_2024_08_06
GPT_4O_2024_11_20
GPT_4O_MINI_2024_07_18 (default)
O3_MINI_2025_01_31
TEXT_DAVINCI_003

Basic Usage¶

from uipath_llamaindex.llms import UiPathOpenAI
from llama_index.core.llms import ChatMessage

# Create an LLM instance with default settings
llm = UiPathOpenAI()

# Create chat messages
messages = [
    ChatMessage(
        role="system", content="You are a pirate with colorful personality."
    ),
    ChatMessage(role="user", content="Hello"),
]

# Generate a response
response = llm.chat(messages)
print(response)

Custom Model Configuration¶

from uipath_llamaindex.llms import UiPathOpenAI, OpenAIModel

# Use a specific model
llm = UiPathOpenAI(model=OpenAIModel.GPT_4O_2024_11_20)

# Or use a model string directly
llm = UiPathOpenAI(model="gpt-4o-2024-11-20")

UiPathOpenAIEmbedding¶

The UiPathOpenAIEmbedding class provides text embedding capabilities using OpenAI's embedding models through UiPath.

Available Embedding Models¶

The following embedding models are available through the OpenAIEmbeddingModel enum:

TEXT_EMBEDDING_ADA_002 (default)
TEXT_EMBEDDING_3_LARGE

Basic Usage¶

from uipath_llamaindex.embeddings import UiPathOpenAIEmbedding

# Create an embedding model instance
embed_model = UiPathOpenAIEmbedding()

# Get embeddings for a single text
result = embed_model.get_text_embedding("the quick brown fox jumps over the lazy dog")
print(f"Embedding dimension: {len(result)}")

Batch Embeddings¶

from uipath_llamaindex.embeddings import UiPathOpenAIEmbedding

embed_model = UiPathOpenAIEmbedding()

# Get embeddings for multiple texts
texts = [
    "Hello world",
    "How are you?",
    "This is a test"
]

embeddings = embed_model.get_text_embedding_batch(texts)
print(f"Number of embeddings: {len(embeddings)}")

Integration with LlamaIndex¶

Both classes integrate seamlessly with LlamaIndex components:

Using with Agents¶

from llama_index.core.agent import ReActAgent
from llama_index.core.tools import FunctionTool
from uipath_llamaindex.llms import UiPathOpenAI

def multiply(a: int, b: int) -> int:
    """Multiply two integers and returns the result."""
    return a * b

multiply_tool = FunctionTool.from_defaults(fn=multiply)

# Create agent with UiPath LLM
agent = ReActAgent.from_tools(
    [multiply_tool], 
    llm=UiPathOpenAI(model=OpenAIModel.GPT_4O_2024_11_20)
)

response = agent.chat("What is 21 multiplied by 2?")

Using with VectorStoreIndex¶

from llama_index.core import VectorStoreIndex, Document
from uipath_llamaindex.llms import UiPathOpenAI
from uipath_llamaindex.embeddings import UiPathOpenAIEmbedding

# Create documents
documents = [
    Document(text="This is a sample document about artificial intelligence."),
    Document(text="Machine learning is a subset of AI that focuses on algorithms."),
]

# Create index with UiPath models
index = VectorStoreIndex.from_documents(
    documents,
    embed_model=UiPathOpenAIEmbedding()
)

# Create query engine with UiPath LLM
query_engine = index.as_query_engine(
    llm=UiPathOpenAI(model=OpenAIModel.GPT_4O_2024_11_20)
)

response = query_engine.query("What is machine learning?")