How to Set Up the OpenEnv Web Interface for Interactive Debugging

Setting up the OpenEnv web interface requires setting the ENABLE_WEB_INTERFACE=true environment variable before launching the environment, which serves a Gradio-based UI at http://localhost:8000/web for real-time debugging and interaction.

The OpenEnv web interface provides a browser-based debugging environment for exploring observations, sending actions, and visualizing state without writing additional client code. According to the huggingface/OpenEnv source code, this feature is conditionally compiled into the server and automatically routes to /web when enabled.

Enabling the Web Interface

The web UI is disabled by default to minimize overhead during API-only deployments. Activating it requires a single environment variable before the server initializes.

Environment Variable Configuration

Set ENABLE_WEB_INTERFACE=true in your shell or container environment. This flag is respected across all official environment templates including wildfire, textarena, snake, and openapp.

In envs/wildfire_env/README.md, the configuration is documented as:

export ENABLE_WEB_INTERFACE=true

Similarly, envs/textarena_env/README.md and envs/snake_env/README.md confirm this pattern activates the Gradio interface at the /web endpoint.

Accessing the Interface

Once the environment launches with the flag set, navigate to:

http://localhost:8000/web

The interface loads immediately, allowing you to inspect observations and replay actions interactively.

Deployment Methods

You can enable the OpenEnv web interface through multiple deployment paths depending on your infrastructure.

Local Development Setup

For local debugging, export the variable and run via the CLI:

export ENABLE_WEB_INTERFACE=true
openenv run my_env

The server now exposes the UI alongside the standard API.

Docker Deployment

When building containerized environments, pass the flag as a build argument or runtime environment variable:

docker build -t my-env \
  --build-arg ENABLE_WEB_INTERFACE=true \
  -f Dockerfile .

docker run -p 8000:8000 my-env

The container serves the web interface on port 8000 as specified in the image configuration.

API-Only Mode

To disable the interface and run a headless server, use the --no-interface flag parsed by src/openenv/cli/cli.py:

openenv run my_env --no-interface

Alternatively, explicitly set ENABLE_WEB_INTERFACE=false to prevent the Gradio server from mounting.

Supported Environment Templates

The ENABLE_WEB_INTERFACE flag is standardized across the repository. Documentation in docs/source/environments/textarena.md confirms that TextArena environments expose the Gradio UI under /web when the variable is present. This consistency extends to:

• Wildfire: Documented in envs/wildfire_env/README.md
• TextArena: Documented in envs/textarena_env/README.md and docs/source/environments/textarena.md
• Snake: Documented in envs/snake_env/README.md

Each template routes observations and action inputs through the same web component, ensuring a uniform debugging experience regardless of the specific domain.

Implementation Details

The CLI entry point in src/openenv/cli/cli.py handles the conditional logic for UI activation. When ENABLE_WEB_INTERFACE is detected, the server mounts the Gradio application at the /web route during startup. If the --no-interface argument is passed, the CLI overrides the environment variable and skips the mounting process, resulting in a lightweight API server suitable for production training pipelines.

Summary

• Set ENABLE_WEB_INTERFACE=true before launching to activate the debugging UI.
• Access the interface at http://localhost:8000/web once the server starts.
• Use --no-interface or set the variable to false for API-only deployments.
• Reference template documentation in envs/*/README.md files for environment-specific confirmation.
• Implementation resides in src/openenv/cli/cli.py, which parses both the environment variable and CLI flag.

Frequently Asked Questions

How do I disable the OpenEnv web interface when running in production?

Pass the --no-interface flag to the openenv run command or set ENABLE_WEB_INTERFACE=false in your environment. As implemented in src/openenv/cli/cli.py, this prevents the Gradio server from mounting, reducing memory overhead and preventing unauthorized browser access.

Does the web interface work with all OpenEnv environments?

Yes, the flag is standardized across all official templates. Documentation in envs/wildfire_env/README.md, envs/textarena_env/README.md, and envs/snake_env/README.md confirms that setting ENABLE_WEB_INTERFACE=true enables the interactive UI for Wildfire, TextArena, and Snake environments respectively.

What technology powers the OpenEnv web interface?

The interface is built on Gradio, a Python library for creating customizable UI components. When enabled, the framework automatically serves the components at the /web endpoint, allowing you to visualize environment states and submit actions through form inputs rather than raw API calls.

Can I change the default port for the web interface?

The examples use port 8000, which is the default server port. You can map this to any host port using Docker's -p flag (e.g., -p 3000:8000) or configure your reverse proxy to forward traffic to the internal 8000 port where the Gradio UI is mounted.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}