LangSmith — Animate | AI Wisdom

← Back to topic
🎞 ANIMATE · Mode 3 of 5 · +100 XP— LangSmith Tracing
📋Config
⚙️Decorate
🔄Execute
✅Query
🎯 Next: Test →
Step 1 / 4
Speed
Configure Tracing
Set LANGCHAIN_TRACING_V2 env vars to enable automatic tracing💡 Analogy
Like turning on flight recorder mode — once enabled, everything is automatically logged.
Three env vars enable LangSmith. TRACING_V2=true turns on tracing globally. API_KEY authenticates to LangSmith. PROJECT groups runs.
Console Output
> LANGCHAIN_TRACING_V2: true
> LANGCHAIN_PROJECT: my-rag-app
> Tracing active: all LangChain calls will be uploaded
●langsmith_trace.py
Configure TracingL6–8
1import os
2from langsmith import Client, traceable
3from langchain_openai import ChatOpenAI
4from langchain_core.prompts import ChatPromptTemplate
5 
6os.environ["LANGCHAIN_TRACING_V2"] = "true"
7os.environ["LANGCHAIN_API_KEY"] = "<your-key>"
8os.environ["LANGCHAIN_PROJECT"] = "my-rag-app"
9 
10llm = ChatOpenAI(model="gpt-4o-mini", temperature=0)
11 
12prompt = ChatPromptTemplate.from_messages([
13    ("system", "Answer concisely."),
14    ("human", "{question}"),
15])
16 
17chain = prompt | llm
18 
19@traceable(name="rag-pipeline", tags=["production"])
20def rag_pipeline(question: str, context: str) -> str:
21    response = chain.invoke({
22        "question": f"Context: {context}\n\nQ: {question}"
23    })
24    return response.content
25 
26# All calls are traced to LangSmith automatically
27answer = rag_pipeline(
28    question="What is RAG?",
29    context="RAG retrieves relevant docs before generating answers."
30)
31 
32# Query runs programmatically
33client = Client()
34runs = client.list_runs(project_name="my-rag-app", limit=10)
35for run in runs:
36    print(f"{run.name}: {run.total_tokens} tokens, {run.total_cost:.4f} USD")
36 lines
Timeline — 4 Steps
Configure Tracing
Set LANGCHAIN_TRACING_V2 env vars to enable automatic tracing💡 Analogy
Like turning on flight recorder mode — once enabled, everything is automatically logged.
Three env vars enable LangSmith. TRACING_V2=true turns on tracing globally. API_KEY authenticates to LangSmith. PROJECT groups runs.
Console Output
> LANGCHAIN_TRACING_V2: true
> LANGCHAIN_PROJECT: my-rag-app
> Tracing active: all LangChain calls will be uploaded

Was this page helpful?

Discussion

Set LANGCHAIN_TRACING_V2 env vars to enable automatic tracing

💡 Analogy

Like turning on flight recorder mode — once enabled, everything is automatically logged.

Three env vars enable LangSmith. TRACING_V2=true turns on tracing globally. API_KEY authenticates to LangSmith. PROJECT groups runs.

Console Output

> LANGCHAIN_TRACING_V2: true
> LANGCHAIN_PROJECT: my-rag-app
> Tracing active: all LangChain calls will be uploaded

Set LANGCHAIN_TRACING_V2 env vars to enable automatic tracing

💡 Analogy

Like turning on flight recorder mode — once enabled, everything is automatically logged.

Three env vars enable LangSmith. TRACING_V2=true turns on tracing globally. API_KEY authenticates to LangSmith. PROJECT groups runs.

Console Output

> LANGCHAIN_TRACING_V2: true
> LANGCHAIN_PROJECT: my-rag-app
> Tracing active: all LangChain calls will be uploaded