LLM Model Configuration By Environment Variable

Carry Code supports configuring AI models through environment variables, enabling seamless integration with CI/CD pipelines, containerized deployments, and existing development workflows.

---

Overview

Environment variable configuration allows you to:

• Quick Setup: Configure models without editing config files
• CI/CD Integration: Use secrets management in pipelines
• Container Deployment: Inject credentials via environment variables
• Multi-Environment: Different models for different environments

Priority

Environment variable configurations have higher priority than config file settings. When both are present, environment variables take precedence.

---

Supported Environment Variables

Provider-Specific Variables

Carry Code automatically detects and configures models from the following environment variables:

Provider	API Key Variable	Base URL Variable (Optional)	Model Variable (Optional)
Anthropic	ANTHROPIC_AUTH_TOKEN	ANTHROPIC_BASE_URL	ANTHROPIC_MODEL
OpenAI	OPENAI_API_KEY	OPENAI_BASE_URL	OPENAI_MODEL
Google Gemini	GEMINI_API_KEY	GOOGLE_GEMINI_BASE_URL	GEMINI_MODEL
OpenRouter	OPENROUTER_API_KEY	OPENROUTER_BASE_URL	OPENROUTER_MODEL

Global Default Model

Variable	Description	Example
CARRYCODE_MODEL	Set the default model to use	anthropic-env:claude-opus-4-6 or gpt-4o

---

Usage Examples

Basic Setup

Set the API key for your preferred provider:

# Anthropic Claude
export ANTHROPIC_AUTH_TOKEN="your-anthropic-api-key"

# OpenAI
export OPENAI_API_KEY="your-openai-api-key"

# Google Gemini
export GEMINI_API_KEY="your-gemini-api-key"

# OpenRouter
export OPENROUTER_API_KEY="your-openrouter-api-key"

Custom Base URL

Override the default API endpoint:

# Use a custom Anthropic endpoint
export ANTHROPIC_AUTH_TOKEN="your-key"
export ANTHROPIC_BASE_URL="https://your-proxy.example.com"

# Use a custom OpenAI endpoint (e.g., Azure OpenAI)
export OPENAI_API_KEY="your-azure-key"
export OPENAI_BASE_URL="https://your-resource.openai.azure.com/openai/deployments/your-deployment"

Specify Model

Override the default model for a provider:

# Use Claude Opus 4.5
export ANTHROPIC_AUTH_TOKEN="your-key"
export ANTHROPIC_MODEL="claude-opus-4-5-20250514"

# Use GPT-4o mini
export OPENAI_API_KEY="your-key"
export OPENAI_MODEL="gpt-4o-mini"

# Use Gemini 2.5 Pro
export GEMINI_API_KEY="your-key"
export GEMINI_MODEL="gemini-2.5-pro"

Set Default Model

Use CARRYCODE_MODEL to specify which model to use by default:

# Use Anthropic model configured via environment variable
export ANTHROPIC_AUTH_TOKEN="your-key"
export CARRYCODE_MODEL="anthropic-env:claude-opus-4-6"

# Or use a model name directly (matches first provider with that model)
export OPENAI_API_KEY="your-key"
export OPENAI_MODEL="gpt-4o"
export CARRYCODE_MODEL="gpt-4o"

---

Provider ID Naming

When models are configured via environment variables, Carry Code creates provider entries with the ID format {provider}-env:

Environment Variable	Generated Provider ID
ANTHROPIC_AUTH_TOKEN	anthropic-env
OPENAI_API_KEY	openai-env
GEMINI_API_KEY	gemini-env
OPENROUTER_API_KEY	openrouter-env

This allows you to reference these models in commands:

# Switch to environment-configured Anthropic model
/model switch anthropic-env

# Set as default
/model default openai-env

---

Configuration Inheritance

Models configured via environment variables inherit default parameters from Carry Code’s built-in provider presets:

• Context Window: Maximum context length
• Temperature: Default sampling temperature
• Top-P: Nucleus sampling parameter
• Max Tokens: Maximum output tokens

These defaults are defined in ConfigProviders.toml and are automatically applied when you configure a model via environment variables.

---

Shell Configuration

Bash

Add to ~/.bashrc or ~/.bash_profile:

# Carry Code Model Configuration
export ANTHROPIC_AUTH_TOKEN="your-anthropic-key"
export OPENAI_API_KEY="your-openai-key"
export CARRYCODE_MODEL="anthropic-env:claude-opus-4-6"

Zsh

Add to ~/.zshrc:

# Carry Code Model Configuration
export ANTHROPIC_AUTH_TOKEN="your-anthropic-key"
export OPENAI_API_KEY="your-openai-key"
export CARRYCODE_MODEL="anthropic-env:claude-opus-4-6"

Fish

Add to ~/.config/fish/config.fish:

# Carry Code Model Configuration
set -gx ANTHROPIC_AUTH_TOKEN "your-anthropic-key"
set -gx OPENAI_API_KEY "your-openai-key"
set -gx CARRYCODE_MODEL "anthropic-env:claude-opus-4-6"

---

Docker & Kubernetes

Docker

Pass environment variables when running a container:

docker run -it \
  -e ANTHROPIC_AUTH_TOKEN="your-key" \
  -e CARRYCODE_MODEL="anthropic-env:claude-opus-4-6" \
  carrycode/carrycode:latest

Or use an environment file:

# .env file
ANTHROPIC_AUTH_TOKEN=your-key
OPENAI_API_KEY=your-key
CARRYCODE_MODEL=anthropic-env:claude-opus-4-6

docker run -it --env-file .env carrycode/carrycode:latest

Kubernetes

Use ConfigMap or Secret:

apiVersion: v1
kind: Secret
metadata:
  name: carrycode-secrets
type: Opaque
stringData:
  anthropic-auth-token: "your-key"
  openai-api-key: "your-key"
---
apiVersion: v1
kind: Pod
metadata:
  name: carrycode
spec:
  containers:
  - name: carrycode
    image: carrycode/carrycode:latest
    env:
    - name: ANTHROPIC_AUTH_TOKEN
      valueFrom:
        secretKeyRef:
          name: carrycode-secrets
          key: anthropic-auth-token
    - name: CARRYCODE_MODEL
      value: "anthropic-env:claude-opus-4-6"

---

CI/CD Integration

GitHub Actions

Use GitHub Secrets for API keys:

name: AI Code Review
on: [pull_request]

jobs:
  review:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - name: Run CarryCode
        env:
          ANTHROPIC_AUTH_TOKEN: ${{ secrets.ANTHROPIC_API_KEY }}
          CARRYCODE_MODEL: "anthropic-env:claude-opus-4-6"
        run: |
          # Install and run CarryCode
          carrycode "Review the changes in this PR"

GitLab CI

Use GitLab CI/CD variables:

code-review:
  stage: test
  script:
    - carrycode "Review the code changes"
  variables:
    ANTHROPIC_AUTH_TOKEN: $ANTHROPIC_API_KEY
    CARRYCODE_MODEL: "anthropic-env:claude-opus-4-6"

---

Debugging

Verify Environment Variables

Check if your environment variables are set correctly:

# List all Carry Code-related environment variables
env | grep -E "(ANTHROPIC|OPENAI|GEMINI|OPENROUTER|CARRYCODE)"

Check Active Configuration

Within Carry Code, use the /model command to see all configured models, including those from environment variables. Environment-configured models will appear with provider IDs ending in -env.

Common Issues

Issue	Solution
Model not appearing	Ensure the API key variable is set and non-empty
Wrong model selected	Check CARRYCODE_MODEL format matches provider-id:model-name
API errors	Verify the API key is valid and has sufficient credits
Custom URL not working	Ensure base URL includes the full API path

---

Security Best Practices

1. Never commit API keys to version control
2. Use secrets management in CI/CD (GitHub Secrets, GitLab Variables, etc.)
3. Rotate keys regularly for production environments
4. Restrict key permissions when possible (e.g., read-only for CI)
5. Use environment-specific keys for different stages (dev, staging, prod)

---

See Also

• Model Configuration - Configure models via config file
• MCP Servers - Extend AI capabilities
• Getting Started - Quick start guide