Skip to main content
Graphlit can accelerate your Generative AI app development by providing an API-first platform that offers intelligent text extraction, vector-based search, and RAG conversations, along with audio and video transcription support. Langtrace integrates directly with Graphlit, providing comprehensive tracing and monitoring capabilities for your RAG (Retrieval-Augmented Generation) applications. Track everything from data ingestion to conversation completion with detailed performance metrics.

​
Prerequisites

Before you begin, ensure you have the following:
  • An active account on the Graphlit Platform with access to the API Settings page.
  • An API key from Langtrace. Sign up for Langtrace if you haven’t done so already.

​
Setup

  1. Install the Graphlit client and Langtrace’s SDK and initialize the SDK in your code.
Python
# Install the SDK
pip install langtrace-python-sdk graphlit-client
  1. Setup environment variables:
Shell
export LANGTRACE_API_KEY=YOUR_LANGTRACE_API_KEY
export GRAPHLIT_ORGANIZATION_ID=YOUR_GRAPHLIT_ORGANIZATION_ID
export GRAPHLIT_ENVIRONMENT_ID=YOUR_GRAPHLIT_ENVIRONMENT_ID
export GRAPHLIT_JWT_SECRET=YOUR_GRAPHLIT_JWT_SECRET

​
Usage

Initialize Langtrace and Graphlit: 
from langtrace_python_sdk import langtrace  # Must precede other imports
from langtrace_python_sdk.utils.with_root_span import with_langtrace_root_span
from graphlit import Graphlit
from graphlit_api import input_types, enums
from openai import OpenAI

# Initialize Langtrace
langtrace.init()

# Initialize Graphlit and OpenAI clients
graphlit = Graphlit(
    organization_id="YOUR_ORG_ID",
    environment_id="YOUR_ENV_ID",
    jwt_secret="YOUR_JWT_SECRET"
)
client = OpenAI()
Ingest content and create a specification: 
@with_langtrace_root_span()
async def setup_rag():
    # Ingest content
    ingest_response = await graphlit.client.ingest_uri(
        uri="https://example.com",
        is_synchronous=True
    )
    content_id = ingest_response.ingest_uri.id
    
    # Create specification
    specification_input = input_types.SpecificationInput(
        name="OpenAI GPT-4",
        type=enums.SpecificationTypes.COMPLETION,
        serviceType=enums.ModelServiceTypes.OPEN_AI,
        openAI=input_types.OpenAIModelPropertiesInput(
            model=enums.OpenAIModels.GPT4O_128K,
        )
    )
    spec_response = await graphlit.client.create_specification(specification_input)
    return spec_response.create_specification.id
Create and interact with a conversation: 
@with_langtrace_root_span()
async def process_conversation(specification_id, prompt):
    # Create conversation
    conversation_input = input_types.ConversationInput(
        name="RAG Conversation",
        specification=input_types.EntityReferenceInput(
            id=specification_id
        )
    )
    conv_response = await graphlit.client.create_conversation(conversation_input)
    conversation_id = conv_response.create_conversation.id
    
    # Format conversation
    format_response = await graphlit.client.format_conversation(
        prompt,
        conversation_id
    )
    formatted_message = format_response.format_conversation.message.message
    
    # Generate completion using OpenAI
    stream_response = client.chat.completions.create(
        model="gpt-4",
        messages=[{"role": "user", "content": formatted_message}],
        temperature=0.1,
        top_p=0.2,
        stream=True
    )
    
    completion = ""
    for chunk in stream_response:
        if chunk.choices[0].delta.content:
            completion += chunk.choices[0].delta.content
            
    # Store completion
    await graphlit.client.complete_conversation(
        completion,
        conversation_id
    )
    
    return completion
Run the complete RAG pipeline: 
import asyncio

async def main():
    try:
        # Setup RAG environment
        specification_id = await setup_rag()
        
        # Process a question
        response = await process_conversation(
            specification_id,
            "What are the key points from the document?"
        )
        print(response)
        
    except Exception as e:
        print(f"Error: {e}")

# Run the pipeline
asyncio.run(main())
You can now view your traces on the Langtrace dashboard, including:
  • Data ingestion metrics
  • Specification creation details
  • Conversation flow and completion times
  • Error tracking and diagnostics
traces

​
Additional Resources

⌘I