ragmetrics-pkg 0.1.4

A package for integrating RagMetrics with LLM calls

Ragmetrics

Call all LLM APIs using the OpenAI format.

Ragmetrics manages:

• Translate inputs to provider's completion endpoints
• Consistent Traces logged on Ragmetrics portal
• Realtime Monitoring and AB Test Evaluations

Support for more providers. Contact our team to get it done.

Usage Docs

pip install ragmetrics-pkg

Portal Login

import ragmetrics

## login to ragmetrics account via portal key
ragmetrics.login(key="febfe*****************", off=False)

Here off flag helps developer to toggle traces.
off -> True (will turn off the traces and vice versa)

Monitoring/AB Testing/Evaluation (based on compare models)

import ragmetrics

## starts monitoring all LLM calls from this client
ragmetrics.monitor(client, context={"user": True})

Context will help to feed as dataset over Ragmetrics Tasks for Evaluation

Code Utilisation

from openai import OpenAI
import ragmetrics

client = OpenAI()

# Start monitoring all LLM calls from this client
ragmetrics.monitor(client, context={"user": True})

# Use regular OpenAI API calls with an extra optional metadata parameter
chat_completion = client.chat.completions.create(
  model="gpt-4o", 
  messages=[{"role": "user", "content": "Hello Ragmetrics"}],
  metadata={"pipelineStep":"generation", "property1":"Accuracy and Clarity"},
  comparison_model="gpt-4-turbo"
)

• Here model is to refer the response generation model gpt-4o in our case.
• Message with content for the actual developer's request to AI model.
• Metadata is to drive the experiments as defined below:
  • pipelineStep for selecting the evaluation step as per Ragmetrics platform experiments.
  • property to add multiple criteria needs to be performed under evaluation (will be added with keys property1, property2... etc.)

Code Examples

Example 1

from openai import OpenAI
import ragmetrics

ragmetrics.login(key="febfe*****************", off=False)

client = OpenAI()

ragmetrics.monitor(client, context={"user": True})

chat_completion = client.chat.completions.create(
  model="gpt-4o", 
  messages=[{"role": "user", "content": "Hello World !!!"}],
  metadata={"pipelineStep":"generation", "property1":"Accuracy and Clarity"},
  comparison_model="gpt-4-turbo"
  )

Example 2

from openai import OpenAI
import ragmetrics

ragmetrics.login(key="febfe*****************", off=False)

client = OpenAI()

ragmetrics.monitor(client, context={"user": True})

chat_completion = client.chat.completions.create(
  model="gpt-3.5-turbo", 
  messages=[{"role": "user", "content": "How's the day today?"}],
  metadata={"pipelineStep":"generation", "property1":"Accuracy and Clarity", "property2":"Buddy-Friendly Tone"},
  comparison_model="gpt-4o"
  )

Portal Keys UI on Ragmetrics Portal

Why did we build this

• Need for simplicity: We need to provide developers a hassle-free solution to interact and monitor their interaction with LLM models like Azure, OpenAI, LiteLLM ... etc.

Contributors