LangSmith allows you to:

• Trace every LLM call
• Inspect prompts and responses
• Measure latency
• Debug chains visually
• Monitor local and production apps

LangSmith works even with Ollama and local LLMs. It does not require OpenAI.

1. What LangSmith Tracks in Your App

Once enabled, LangSmith automatically captures:

• Prompt template used
• Input variables (question)
• LLM call (Ollama in your case)
• Model output
• Execution time
• Errors (if any)
• Chain structure (prompt → llm → parser)

You do not need to change your chain logic.

2. Create a LangSmith Account

1. Go to https://smith.langchain.com
2. Sign in using GitHub / Google
3. Create a workspace (default is fine)

After login, you’ll see:

• Dashboard
• Traces
• Projects

3. Get Your LangSmith API Key

1. Open Settings → API Keys
2. Create a new key
3. Copy it (you’ll use it as an environment variable)

4. Enable LangSmith via Environment Variables (IMPORTANT)

LangSmith is enabled entirely via environment variables.

Required Variables

Variable	Purpose
LANGCHAIN_TRACING_V2	Enables tracing
LANGCHAIN_API_KEY	Auth with LangSmith
LANGCHAIN_PROJECT	Groups related traces

Windows (Command Prompt)

set LANGCHAIN_TRACING_V2=true
set LANGCHAIN_API_KEY=your_api_key_here
set LANGCHAIN_PROJECT=ollama-streamlit-demo

5. NO Code Changes Required

Your existing app.py stays exactly the same.

LangChain automatically detects LangSmith when:

• LANGCHAIN_TRACING_V2=true is set
• API key is present

This works because:

• LangChain Core has LangSmith hooks built-in
• Ollama is still treated as an LLM trace

6. Run the App Again

Run using the same command:

python -m streamlit run app.py

Now:

• Ask a question in the UI
• Let the model generate a response

7. View Traces in LangSmith Dashboard

1. Open https://smith.langchain.com
2. Select your project.
3. Click Traces