LangChain

Forward your LangChain application logs directly to Sazabi for comprehensive observability of your LLM-powered applications. Monitor chains, agents, tool calls, and model interactions.

About this data source

LangChain is a framework for building applications with large language models. By configuring OpenTelemetry instrumentation and exporting to Sazabi, you can:

• Trace chain executions through multiple steps
• Monitor agent decisions and tool calls
• Track LLM requests with latency and token metrics
• Detect prompt issues and model errors
• Use AI to analyze and optimize your LLM workflows

Prerequisites

Before you begin, make sure you have:

• A LangChain application (Python, JavaScript, or other supported language)
• OpenTelemetry SDK for your language
• A Sazabi public API key (project-scoped)

Get your API key

Setup

pip install opentelemetry-api opentelemetry-sdk opentelemetry-exporter-otlp

npm install @opentelemetry/api @opentelemetry/sdk-node @opentelemetry/exporter-trace-otlp-http

OTEL_EXPORTER_OTLP_ENDPOINT=https://otlp.<region>.intake.sazabi.com
OTEL_EXPORTER_OTLP_HEADERS="Authorization=Bearer <your-api-key>"
OTEL_SERVICE_NAME=my-langchain-app

from opentelemetry import trace
from opentelemetry.sdk.trace import TracerProvider
from opentelemetry.sdk.trace.export import BatchSpanProcessor
from opentelemetry.exporter.otlp.proto.http.trace_exporter import OTLPSpanExporter

# Configure OTLP exporter (uses env vars automatically)
exporter = OTLPSpanExporter()
provider = TracerProvider()
processor = BatchSpanProcessor(exporter)
provider.add_span_processor(processor)
trace.set_tracer_provider(provider)

const { NodeSDK } = require("@opentelemetry/sdk-node");
const { OTLPTraceExporter } = require("@opentelemetry/exporter-trace-otlp-http");

const sdk = new NodeSDK({
  traceExporter: new OTLPTraceExporter(),
  // Uses env vars for configuration
});

sdk.start();

What gets captured

LangChain with OpenTelemetry captures:

• Chain executions: Start, completion, and errors for each chain
• Agent traces: Agent decisions, tool selections, and reasoning steps
• LLM calls: Model requests, responses, token counts, and latencies
• Tool invocations: Tool calls with inputs and outputs
• Retriever queries: Document retrieval operations

Features

• Real-time tracing of chains, agents, and tool calls via OpenTelemetry
• LLM call monitoring with token usage and latency metrics
• Compatible with LangGraph and other LangChain integrations
• Works with any language that has an OpenTelemetry SDK

Verifying traces are flowing

Once configured, verify that traces are flowing to Sazabi:

1. Run a chain: Execute one of your LangChain chains or agents.

2. Ask the assistant: Open a thread in Sazabi and ask "Show me recent LangChain traces" or "What LLM calls were made in the last hour?"

Troubleshooting