Monitor AI chat interactions with Gemini 2.5 and Langfuse tracing
This workflow contains community nodes that are only compatible with the self-hosted version of n8n.
# How it works
This workflow is a simple AI Agent that connects to Langfuse so send tracing data to help monitor LLM interactions.
The main idea is to create a custom LLM model that allows the configuration of callbacks, which are used by langchain to connect applications such Langfuse.
This is achieves by using the "langchain code" node:
- Connects a LLM model sub-node to obtain the model variables (model name, temp and provider) - Creates a generic langchain initChatModel with the model parameters.
- Return the LLM to be used by the AI Agent node.
## 📋 Prerequisites
- Langfuse instance (cloud or self-hosted) with API credentials
- LLM API key (Gemini, OpenAI, Anthropic, etc.)
- n8n >= 1.98.0 (required for LangChain code node support in AI Agent)
## ⚙️ Setup
1. Add these to your n8n instance:
```bash
# Langfuse configuration
LANGFUSE_SECRET_KEY=your_secret_key
LANGFUSE_PUBLIC_KEY=your_public_key
LANGFUSE_BASEURL=https://cloud.langfuse.com # or your self-hosted URL
# LLM API key (example for Gemini)
GOOGLE_API_KEY=your_api_key
```
Alternative: Configure these directly in the LangChain code node if you prefer not to use environment variables
2. Import the workflow JSON
3. Connect your preferred LLM model node
4. Send a test message to verify tracing appears in Langfuse
