campers 0.1.0

CLI tool for managing remote development environments on the cloud

Campers

Local development experience. Remote cloud resources.

Campers is a command-line tool that manages disposable remote development environments on the cloud (currently AWS EC2). It allows you to offload heavy computation to the cloud while keeping your local development workflow intact.

It bridges the gap between your local machine and a cloud instance by handling provisioning, file synchronization, and network tunneling automatically.

How you use it

The goal of Campers is to make a remote cloud instance feel like localhost.

Imagine you are working on a resource-intensive project, like a large microservices stack or a deep learning model. Your local machine is struggling with heat and memory limits.

With Campers, the workflow looks like this:

1. Configuration: You add a campers.yml file to your project root. This defines the hardware you need (e.g., an instance type like p3.2xlarge), the setup steps, and which ports to forward.

2. Spin Up: You run campers run in your terminal. In the background, Campers provisions the instance, configures it (via shell scripts or Ansible), and establishes a real-time, two-way file sync using Mutagen.

3. Development: You stay on your laptop.

   • You edit code in your local editor (VS Code, Vim, etc.). Changes are synced instantly to the cloud instance.
   • You run your application on the remote instance.
   • You access the application via localhost in your browser. Campers tunnels the traffic through SSH automatically.

4. Exit Options: When you press Q or Ctrl+C, you'll be prompted to choose:

   • Stop (default): Instance is stopped but preserved. Resume later with campers run.
   • Keep running: Disconnect locally but keep the instance running. Ideal for demos where clients need continued access.
   • Destroy: Terminate and delete everything. You can also run campers destroy anytime.

Use Cases

Data Science & Pipelines Ideal for ad-hoc data science projects. Run resource-intensive data pipelines or train models on high-end cloud hardware.

It also solves data residency challenges. Many organizations strictly prohibit storing PII on developer laptops. By spinning up a camp in a compliant cloud region, you can develop against real datasets without ever downloading sensitive data to your local machine.

Isolated Environments Instead of cluttering your local machine with databases and system dependencies, you can define a clean, reproducible environment for each project. If the environment breaks, you simply destroy it and create a new one.

Heavy Compilation If you are compiling large C++ or Rust projects, you can provision a high-core instance (like a c6a.24xlarge) for the duration of the build. You get the build speed of a workstation without maintaining the hardware.

Client Demos Share running applications with clients by exposing ports publicly. Use public_ports to open security group rules, then select "Keep running" on exit so clients can continue accessing your demo while you disconnect.

Features

• Mutagen Sync: Uses Mutagen for high-performance file synchronization. It is not rsync; it uses a real-time, bi-directional sync agent that is orders of magnitude faster for large projects (like node_modules).
• Automatic Port Forwarding: Tunnels remote ports to your local machine based on your configuration.
• Public Port Exposure: Open ports directly for external access - perfect for client demos.
• Ansible Integration: Supports running Ansible playbooks to configure the instance on startup.
• Cost Control: Encourages an ephemeral workflow where instances are destroyed when not in use.
• TUI Dashboard: A terminal interface to monitor logs, sync status, and instance health.

Simple Configuration

Campers uses a single YAML file to define your infrastructure and provisioning. Here is a complete example:

# campers.yml

# Define reusable variables to keep config clean
vars:
  project_name: my-ml-project
  # Use standard linux paths
  remote_path: /home/ubuntu/${project_name}

# Define reusable Ansible playbooks (idempotent setup)
playbooks:
  python-setup:
    - name: Install Python Tools
      hosts: all
      tasks:
        - pip: {name: [numpy, pandas, jupyter], state: present}

  deep-learning:
    - name: Install PyTorch & TensorBoard
      hosts: all
      tasks:
        - pip: {name: [torch, torchvision, tensorboard], state: present}

# Define your camps (machines)
camps:
  # 1. Cheap dev environment
  dev:
    instance_type: t3.medium
    # Uses variable defined above
    command: cd ${remote_path} && bash

  # 2. Interactive experimentation (Jupyter)
  experiment:
    instance_type: g4dn.xlarge
    # Use the Deep Learning AMI
    ami:
      query:
        name: "Deep Learning Base AMI (Ubuntu*)*"
        owner: "amazon"
    # Open Jupyter on your laptop's localhost:8888
    ports: [8888]
    ansible_playbooks: [python-setup]
    command: jupyter lab --ip=0.0.0.0 --port=8888

  # 3. Heavy training job (TensorBoard)
  training:
    instance_type: p3.2xlarge
    # Forward TensorBoard to localhost:6006
    ports: [6006]
    ansible_playbooks: [python-setup, deep-learning]

    # Run every time the instance starts (e.g., pull latest data)
    startup_script: |
      cd ${remote_path}
      dvc pull data/

    # Run background monitoring and main training script
    command: |
      cd ${remote_path}
      tensorboard --logdir logs --port 6006 &
      python train_model.py

  # 4. Client demo (publicly accessible)
  demo:
    instance_type: t3.medium
    # Open ports for external access (clients can hit the public IP)
    public_ports: [80, 3000]
    command: npm start

How you use them:

# Start the cheap coding environment
campers run dev

# Switch to the GPU machine for notebooks
campers run experiment

# Launch the heavy training job
campers run training

# Start a client demo (share the public IP with clients)
campers run demo

# Check status of all your camps (showing estimated monthly costs)
campers list
# ┏━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━┓
# ┃ NAME               ┃ INSTANCE-ID  ┃ STATUS     ┃ REGION         ┃ COST/MONTH           ┃
# ┡━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━┩
# │ campers-dev        │ i-0abc123def │ running    │ us-east-1      │ $29.95/month         │
# │ campers-experiment │ i-0def456abc │ stopped    │ us-east-1      │ $4.00/month          │
# └────────────────────┴──────────────┴────────────┴────────────────┴──────────────────────┘

Full Control

Since you get a standard Linux instance, you can run multiple services at once. You might use supervisord or docker compose to spin up Jupyter, TensorBoard, and a database simultaneously. Campers will automatically forward all the ports you specify.

Environment Forwarding

Campers securely forwards your local environment variables (like API keys) to the remote instance. You can configure exactly which variables to send using regex filters:

defaults:
  # Only forward specific safe variables
  env_filter:
    - ^AWS_.*
    - ^WANDB_API_KEY

Quick Start

# Install via pip
pip install campers

# Or run instantly with uv (recommended)
uvx campers run

# Initialize a configuration in your current directory
campers init

# Spin up your camp
campers run

Documentation

Full documentation is available at kamilc.github.io/campers

• Getting Started
• Configuration Reference
• CLI Commands
• Examples

License

MIT