Skip to main content

Configuring Your Agent

Now that you've created and run your first agent, let's learn how to configure it with environment variables. Everything in this chapter happens in the chatbot-agent project from the previous chapter.

What You'll Learn​

  • Using .env files for configuration
  • Accessing environment variables in agent code
  • Documenting configuration with .env.example

The .env File​

AgentVisor™ supports .env files to pass configuration and secrets to your agent. Create a .env file in your agent directory:

chatbot-agent/
├── agent.py
├── langgraph.json
├── requirements.txt
├── ...
└── .env              ← auto-discovered

Basic Example​

# chatbot-agent/.env

# Which Ollama model the chatbot uses (agent.py reads OLLAMA_MODEL)
OLLAMA_MODEL=llama3.2:1b

# API keys for hosted providers, should you switch the agent to one
# OPENAI_API_KEY=sk-...

# Custom configuration
GREETING_PREFIX=AgentVisor says:

When you run agentvisor serve, the .env file is automatically discovered and loaded:

agentvisor serve .
# Output: Found .env file: /path/to/chatbot-agent/.env

Reading Variables in Your Agent​

Access environment variables with standard Python os.environ:

import os

# agent.py already does this for the model name:
OLLAMA_MODEL = os.environ.get("OLLAMA_MODEL", "llama3.2:1b")

def chat(state):
    prefix = os.environ.get("GREETING_PREFIX", "Response:")
    api_key = os.environ.get("OPENAI_API_KEY")
    # ...

LangChain libraries automatically read common variables like OPENAI_API_KEY — no extra code needed.

Documenting with .env.example​

Create a .env.example file (without real secrets) to document required variables for your team:

# .env.example (committed to git)
OPENAI_API_KEY=your-key-here
GREETING_PREFIX=Hello:
tip

Add .env to your .gitignore to avoid committing secrets. The .env.example file serves as documentation for what variables are needed.

Next Steps​