Observability for LangChain with Grafana Tempo

Step 1

Install the Traceloop SDK and initialize it. It will automatically create traces from all the chains you've created through LangChain, with prompts and completions as separate spans.

from langchain.schema import SystemMessage, HumanMessage
from langchain.prompts import ChatPromptTemplate, HumanMessagePromptTemplate
from langchain_community.chat_models import ChatOpenAI
from langchain.chains import LLMChain, SequentialChain, TransformChain

from traceloop.sdk import Traceloop

Traceloop.init(app_name="langchain_example")

chat = ChatOpenAI(temperature=0)

transform = TransformChain(
    input_variables=["subject"],
    output_variables=["prompt"],
    transform=lambda subject: {"prompt": f"Tell me a joke about {subject}."},
)

first_prompt_messages = [
    SystemMessage(content="You are a funny sarcastic nerd."),
    HumanMessage(content="{prompt}"),
]
first_prompt_template = ChatPromptTemplate.from_messages(first_prompt_messages)
first_chain = LLMChain(llm=chat, prompt=first_prompt_template, output_key="joke")

second_prompt_messages = [
    SystemMessage(content="You are an Elf."),
    HumanMessagePromptTemplate.from_template(
        "Translate the joke below into Sindarin language:\n {joke}"
    ),
]
second_prompt_template = ChatPromptTemplate.from_messages(second_prompt_messages)
second_chain = LLMChain(llm=chat, prompt=second_prompt_template)

workflow = SequentialChain(
    chains=[transform, first_chain, second_chain], input_variables=["subject"]
)
print(workflow({"subject": "OpenTelemetry"}))


import * as traceloop from "@traceloop/node-server-sdk";

import { createOpenAIToolsAgent, AgentExecutor } from "langchain/agents";
import { SerpAPI } from "@langchain/community/tools/serpapi";
import { Calculator } from "langchain/tools/calculator";
import { ChatOpenAI } from "@langchain/openai";
import { pull } from "langchain/hub";

traceloop.initialize({
  appName: "sample_langchain",
  apiKey: process.env.TRACELOOP_API_KEY,
  disableBatch: true,
});

const llm = new ChatOpenAI({});
const tools = [new Calculator(), new SerpAPI()];
const prompt = await pull(
  "hwchase17/openai-tools-agent",
);
const agent = await createOpenAIToolsAgent({ llm, tools, prompt });
const agentExecutor = new AgentExecutor({
  agent,
  tools,
});
const result = await agentExecutor.invoke({
  input:
    "By searching the Internet, find how many albums " + 
    "has Boldy James dropped since 2010 and how many " + 
    "albums has Nas dropped since 2010? Find who dropped " + 
    "more albums and show the difference in percent.",
});
return result;

Step 2

Make sure you have a Grafana Agent installed and running in your system. Go to the Grafana Cloud account page under https://grafana.com/orgs/<your org name>, and click on Send Traces under Tempo. In Grafana Data Source settings, note the URL value. Click Generate now to generate an API key and copy it. Note also the Stack ID value (you can find it in the URL https://grafana.com/orgs/<Your Org Name>/stacks/<Stack ID>). Add this to the configuration of your Grafana Agent.

TRACELOOP_BASE_URL=https://api.honeycomb.io
TRACELOOP_HEADERS="x-honeycomb-team=<YOUR_API_KEY>"

Docs

Observability for LangChain with Grafana Tempo

Step 1

Step 2

Discover use cases

Trace chains built with LangChain

More things you can connect to {platform}

Chroma

LlamaIndex

Gemini

together.ai

Ollama

Bedrock

Watsonx

Cohere

Anthropic

Mistral AI

OpenAI