LLamaIndex

The way we recommend instrumenting LlamaIndex is to use the OpenTelemetry specific instrumentation provided by OpenLLMetry: opentelemetry-instrumentation-llamaindex.

Installation¶

Install the opentelemetry-instrumentation-llamaindex package:


pip install opentelemetry-instrumentation-llamaindex

Usage¶

Let's use LlamaIndex with OpenAI as an example.

You only need to include the LlamaIndexInstrumentor and call its instrument method to enable the instrumentation.


import logfire
from llama_index.core import VectorStoreIndex
from llama_index.llms.openai import OpenAI
from llama_index.readers.web import SimpleWebPageReader
from opentelemetry.instrumentation.llamaindex import LlamaIndexInstrumentor

logfire.configure()
LlamaIndexInstrumentor().instrument()

# URL for Pydantic's main concepts page
url = 'https://docs.pydantic.dev/latest/concepts/models/'

# Load the webpage
documents = SimpleWebPageReader(html_to_text=True).load_data([url])

# Create index from documents
index = VectorStoreIndex.from_documents(documents)

# Initialize the LLM
query_engine = index.as_query_engine(llm=OpenAI())

# Get response
response = query_engine.query('Can I use RootModels without subclassing them? Show me an example.')
print(str(response))
"""
Yes, you can use RootModels without subclassing them. Here is an example:

```python
from pydantic import RootModel

Pets = RootModel[list[str]]

my_pets = Pets.model_validate(['dog', 'cat'])

print(my_pets[0])
#> dog
print([pet for pet in my_pets])
#> ['dog', 'cat']
"""

Instrument the underlying LLM¶

The LlamaIndexInstrumentor will specifically instrument the LlamaIndex library, not the LLM itself. If you want to instrument the LLM, you'll need to instrument it separately:

For OpenAI, you can use the OpenAI, you can check the OpenAI documentation.
For Anthropic, you can check the Anthropic documentation.

If you are using a different LLM, and you can't find a way to instrument it, or you need any help, feel free to reach out to us!