diff --git a/docs/source/environments.md b/docs/source/environments.md index febc884bd..05b60f661 100644 --- a/docs/source/environments.md +++ b/docs/source/environments.md @@ -80,6 +80,7 @@ The OpenEnv community has built a catalog of ready-to-run environments that cove

Traffic control scenarios with SUMO simulators for agents that reason about continuous control and scheduling.

📄 Docs + 🤗 HF
@@ -94,7 +95,7 @@ The OpenEnv community has built a catalog of ready-to-run environments that cove

Multi-task text arena for language-game competitions such as Wordle, reasoning puzzles, and program synthesis.

📄 Docs - 🤗 HF + 🤗 HF
@@ -109,7 +110,7 @@ The OpenEnv community has built a catalog of ready-to-run environments that cove

Safety-critical diagnostics from the DIPG benchmark, highlighting guardrails, adversarial prompts, and risk scoring.

📄 Docs - 🤗 HF + 🤗 HF
@@ -117,7 +118,7 @@ The OpenEnv community has built a catalog of ready-to-run environments that cove

Classic snake game environment for RL research with configurable grids, partial observability, and customizable rewards.

📄 Docs - 🤗 HF + 🤗 HF
@@ -133,7 +134,7 @@ The OpenEnv community has built a catalog of ready-to-run environments that cove

Browser automation environment for web agents with DOM interaction, navigation, and multi-step task completion.

📄 Docs - 🤗 HF + 🤗 HF
@@ -162,7 +163,7 @@ The OpenEnv community has built a catalog of ready-to-run environments that cove

Embodied evaluation environment for testing LLM decision-making in a full 3D driving simulator with irreversible consequences and ethical trolley scenarios.

📄 Docs - 🤗 HF + 🤗 HF
@@ -170,6 +171,7 @@ The OpenEnv community has built a catalog of ready-to-run environments that cove

Chess RL environment powered by the moonfish engine with configurable opponents, position evaluation, and full chess rules.

📄 Docs + 🤗 HF
@@ -205,7 +207,7 @@ The OpenEnv community has built a catalog of ready-to-run environments that cove

Simple 5×5 grid world RL testbed and step-by-step guide for building new OpenEnv environments from scratch.

📄 Docs - 🤗 HF + 🤗 HF
@@ -241,6 +243,7 @@ The OpenEnv community has built a catalog of ready-to-run environments that cove

Python REPL environment for code execution tasks based on the Recursive Language Models paradigm with sandboxed execution and context loading.

📄 Docs + 🤗 HF
@@ -255,6 +258,7 @@ The OpenEnv community has built a catalog of ready-to-run environments that cove

OpenEnv wrapper for Terminal-Bench 2 tasks with local and Docker execution modes for terminal-based agent evaluation.

📄 Docs + 🤗 HF
diff --git a/docs/source/environments/browsergym.md b/docs/source/environments/browsergym.md index d410ef227..3aba295a8 100644 --- a/docs/source/environments/browsergym.md +++ b/docs/source/environments/browsergym.md @@ -7,6 +7,8 @@ BrowserGym is a unified framework for web-based agent tasks that provides access *The `/web` playground on MiniWoB `click-test`: the goal, the page URL and the browser screenshot (screenshots on, 320×240 viewport).* +Try it on the [`openenv/browsergym_env`](https://huggingface.co/spaces/openenv/browsergym_env) Space, or connect a client to `https://openenv-browsergym-env.hf.space`. + ## Why BrowserGym? BrowserGym provides a complete pipeline for developing web agents: train on simple tasks, then evaluate on realistic websites. diff --git a/docs/source/environments/carla.md b/docs/source/environments/carla.md index 17fa0a0b5..ba414a221 100644 --- a/docs/source/environments/carla.md +++ b/docs/source/environments/carla.md @@ -14,6 +14,8 @@ Embodied evaluation environment for testing LLM decision-making in a full 3D dri - Free-roam and maze navigation with configurable traffic - Rubric-based reward system for RL training +The [`openenv/carla_env`](https://huggingface.co/spaces/openenv/carla_env) Space is paused because CARLA needs a GPU. Duplicate it and pick an A10G in the Space settings to run your own copy. + ## What Makes This Different Traditional text benchmarks ask models "what would you do?" This environment shows **what models actually do** when: @@ -48,7 +50,7 @@ with CarlaEnv(base_url="http://localhost:8000").sync() as env: ```bash uv run python examples/carla_env/trolley_problems.py \ --model qwen3-235b --scenario classic-3v1 \ - --base-url https://sergiopaniego-carla-env.hf.space + --base-url https://.hf.space ``` ## Scenarios @@ -156,7 +158,7 @@ result = await env.reset(scenario_config={ ## Examples -The [`examples/carla_env/`](https://github.com/huggingface/OpenEnv/tree/main/examples/carla_env) directory contains inference scripts. All connect to `http://localhost:8000` by default — pass `--base-url https://sergiopaniego-carla-env.hf.space` for the live Space. +The [`examples/carla_env/`](https://github.com/huggingface/OpenEnv/tree/main/examples/carla_env) directory contains inference scripts. All connect to `http://localhost:8000` by default — pass `--base-url https://.hf.space` for your running copy of the Space. ### Trolley Problems @@ -277,7 +279,7 @@ docker build -t carla-env:latest -f server/Dockerfile . docker run --gpus all -p 8000:8000 carla-env:latest ``` -**Live Space**: [sergiopaniego/carla-env](https://huggingface.co/spaces/sergiopaniego/carla-env) +**Space**: [openenv/carla_env](https://huggingface.co/spaces/openenv/carla_env) (paused, duplicate it with an A10G) ### Specifications diff --git a/docs/source/environments/chess.md b/docs/source/environments/chess.md index 97613fc62..d6db25bd5 100644 --- a/docs/source/environments/chess.md +++ b/docs/source/environments/chess.md @@ -7,6 +7,8 @@ A chess reinforcement learning environment for OpenEnv, powered by the [moonfish *The `/web` playground after a few moves: the board and one button per legal move.* +Try it on the [`openenv/chess_env`](https://huggingface.co/spaces/openenv/chess_env) Space, or connect a client to `https://openenv-chess-env.hf.space`. + ## Features - **Full chess rules** via python-chess library diff --git a/docs/source/environments/dipg.md b/docs/source/environments/dipg.md index f80e96f77..4393a0882 100644 --- a/docs/source/environments/dipg.md +++ b/docs/source/environments/dipg.md @@ -1,6 +1,8 @@ # DIPG Safety Environment (DIPGSafetyEnv) +Try it on the [`openenv/dipg_safety_env`](https://huggingface.co/spaces/openenv/dipg_safety_env) Space, or connect a client to `https://openenv-dipg-safety-env.hf.space`. + ## Overview The `DIPGSafetyEnv` is a custom environment built on the OpenEnv framework for Reinforcement Learning research in high-stakes AI safety. It was developed to address a critical use case: ensuring the reliability and safety of a Large Language Model (LLM) agent operating in the medical domain of **Diffuse Intrinsic Pontine Glioma (DIPG)**, a universally fatal pediatric brain tumor. diff --git a/docs/source/environments/grid_world.md b/docs/source/environments/grid_world.md index 2923cbb60..11876d5ad 100644 --- a/docs/source/environments/grid_world.md +++ b/docs/source/environments/grid_world.md @@ -7,6 +7,8 @@ A deterministic 5x5 grid world. The agent starts in the top-left corner and must *The `/web` playground after a few steps: the agent, the goal and one button per direction.* +Try it on the [`openenv/grid_world_env`](https://huggingface.co/spaces/openenv/grid_world_env) Space, or connect a client to `https://openenv-grid-world-env.hf.space`. + ## Quick Start Start the server from `envs/grid_world_env/`: diff --git a/docs/source/environments/repl.md b/docs/source/environments/repl.md index c14955298..052aa4cb3 100644 --- a/docs/source/environments/repl.md +++ b/docs/source/environments/repl.md @@ -15,6 +15,8 @@ The package provides: - `LocalREPLEnv`: the same environment, in process - `LocalRLMRunner`: a local RLM loop that prompts a model, runs its code and handles recursion +Try it on the [`openenv/repl`](https://huggingface.co/spaces/openenv/repl) Space, or connect a client to `https://openenv-repl.hf.space`. + ## Quick Start Start a server: diff --git a/docs/source/environments/snake.md b/docs/source/environments/snake.md index b8273f97f..696b0f955 100644 --- a/docs/source/environments/snake.md +++ b/docs/source/environments/snake.md @@ -7,6 +7,8 @@ A multi-agent snake game environment for OpenEnv, based on [marlenv](https://git *The `/web` playground after a few steps: the snake, the fruit and one button per move.* +Try it on the [`openenv/snake_env`](https://huggingface.co/spaces/openenv/snake_env) Space, or connect a client to `https://openenv-snake-env.hf.space`. + ## Overview The Snake environment wraps the marlenv Snake-v1 environment to provide a clean OpenEnv-compatible interface. Multiple snakes can battle on a fixed size grid map, but this implementation focuses on single-agent gameplay. diff --git a/docs/source/environments/sumo.md b/docs/source/environments/sumo.md index addc1e050..727d1d943 100644 --- a/docs/source/environments/sumo.md +++ b/docs/source/environments/sumo.md @@ -3,6 +3,8 @@ Integration of traffic signal control with the OpenEnv framework via SUMO (Simulation of Urban MObility) and SUMO-RL. +Try it on the [`openenv/sumo_rl_env`](https://huggingface.co/spaces/openenv/sumo_rl_env) Space, or connect a client to `https://openenv-sumo-rl-env.hf.space`. + ## Overview This environment enables reinforcement learning for **traffic signal control** using SUMO, a microscopic traffic simulation package. Train RL agents to optimize traffic light timing and minimize vehicle delays. diff --git a/docs/source/environments/tbench2.md b/docs/source/environments/tbench2.md index 8c642a6c2..ab16a7004 100644 --- a/docs/source/environments/tbench2.md +++ b/docs/source/environments/tbench2.md @@ -8,6 +8,8 @@ OpenEnv wrapper for [Terminal-Bench 2](https://github.com/laude-institute/termin | **Local** | Runs commands in the server process (no Docker) | Hugging Face Spaces, environments without Docker access | | **Docker** | Runs each task in its own container | Full TB2.0 fidelity with custom task images | +Try it on the [`openenv/tbench2`](https://huggingface.co/spaces/openenv/tbench2) Space, or connect a client to `https://openenv-tbench2.hf.space`. + ## Quick Start ```python diff --git a/envs/browsergym_env/README.md b/envs/browsergym_env/README.md index eef7cbbfe..1c40f5df6 100644 --- a/envs/browsergym_env/README.md +++ b/envs/browsergym_env/README.md @@ -22,6 +22,8 @@ BrowserGym is a unified framework for web-based agent tasks that provides access *The `/web` playground on MiniWoB `click-test`: the goal, the page URL and the browser screenshot (screenshots on, 320×240 viewport).* +Try it on the [`openenv/browsergym_env`](https://huggingface.co/spaces/openenv/browsergym_env) Space, or connect a client to `https://openenv-browsergym-env.hf.space`. + ## Why BrowserGym? BrowserGym provides a complete pipeline for developing web agents: train on simple tasks, then evaluate on realistic websites. diff --git a/envs/carla_env/README.md b/envs/carla_env/README.md index e16a57e1a..91d825e45 100644 --- a/envs/carla_env/README.md +++ b/envs/carla_env/README.md @@ -30,6 +30,8 @@ Embodied evaluation environment for testing LLM decision-making in a full 3D dri - Free-roam and maze navigation with configurable traffic - Rubric-based reward system for RL training +The [`openenv/carla_env`](https://huggingface.co/spaces/openenv/carla_env) Space is paused because CARLA needs a GPU. Duplicate it and pick an A10G in the Space settings to run your own copy. + ## What Makes This Different Traditional text benchmarks ask models "what would you do?" This environment shows **what models actually do** when: @@ -64,7 +66,7 @@ with CarlaEnv(base_url="http://localhost:8000").sync() as env: ```bash uv run python examples/carla_env/trolley_problems.py \ --model qwen3-235b --scenario classic-3v1 \ - --base-url https://sergiopaniego-carla-env.hf.space + --base-url https://.hf.space ``` ## Scenarios @@ -172,7 +174,7 @@ result = await env.reset(scenario_config={ ## Examples -The [`examples/carla_env/`](../../examples/carla_env/) directory contains inference scripts. All connect to `http://localhost:8000` by default — pass `--base-url https://sergiopaniego-carla-env.hf.space` for the live Space. +The [`examples/carla_env/`](../../examples/carla_env/) directory contains inference scripts. All connect to `http://localhost:8000` by default — pass `--base-url https://.hf.space` for your running copy of the Space. ### Trolley Problems @@ -293,7 +295,7 @@ docker build -t carla-env:latest -f server/Dockerfile . docker run --gpus all -p 8000:8000 carla-env:latest ``` -**Live Space**: [sergiopaniego/carla-env](https://huggingface.co/spaces/sergiopaniego/carla-env) +**Space**: [openenv/carla_env](https://huggingface.co/spaces/openenv/carla_env) (paused, duplicate it with an A10G) ### Specifications diff --git a/envs/chess_env/README.md b/envs/chess_env/README.md index 89ab7abce..dbecb4d05 100644 --- a/envs/chess_env/README.md +++ b/envs/chess_env/README.md @@ -19,6 +19,8 @@ A chess reinforcement learning environment for OpenEnv, powered by the [moonfish *The `/web` playground after a few moves: the board and one button per legal move.* +Try it on the [`openenv/chess_env`](https://huggingface.co/spaces/openenv/chess_env) Space, or connect a client to `https://openenv-chess-env.hf.space`. + ## Features - **Full chess rules** via python-chess library diff --git a/envs/dipg_safety_env/README.md b/envs/dipg_safety_env/README.md index 980df8fce..250b3d684 100644 --- a/envs/dipg_safety_env/README.md +++ b/envs/dipg_safety_env/README.md @@ -13,6 +13,8 @@ tags: # DIPG Safety Environment (DIPGSafetyEnv) +Try it on the [`openenv/dipg_safety_env`](https://huggingface.co/spaces/openenv/dipg_safety_env) Space, or connect a client to `https://openenv-dipg-safety-env.hf.space`. + ## Overview The `DIPGSafetyEnv` is a custom environment built on the OpenEnv framework for Reinforcement Learning research in high-stakes AI safety. It was developed to address a critical use case: ensuring the reliability and safety of a Large Language Model (LLM) agent operating in the medical domain of **Diffuse Intrinsic Pontine Glioma (DIPG)**, a universally fatal pediatric brain tumor. diff --git a/envs/grid_world_env/README.md b/envs/grid_world_env/README.md index 5af6fc4c4..fb3246209 100644 --- a/envs/grid_world_env/README.md +++ b/envs/grid_world_env/README.md @@ -18,6 +18,8 @@ A deterministic 5x5 grid world. The agent starts in the top-left corner and must *The `/web` playground after a few steps: the agent, the goal and one button per direction.* +Try it on the [`openenv/grid_world_env`](https://huggingface.co/spaces/openenv/grid_world_env) Space, or connect a client to `https://openenv-grid-world-env.hf.space`. + ## Quick Start Start the server from `envs/grid_world_env/`: diff --git a/envs/repl_env/README.md b/envs/repl_env/README.md index f1954bf0e..8b5d7d2bf 100644 --- a/envs/repl_env/README.md +++ b/envs/repl_env/README.md @@ -27,6 +27,8 @@ The package provides: - `LocalREPLEnv`: the same environment, in process - `LocalRLMRunner`: a local RLM loop that prompts a model, runs its code and handles recursion +Try it on the [`openenv/repl`](https://huggingface.co/spaces/openenv/repl) Space, or connect a client to `https://openenv-repl.hf.space`. + ## Quick Start Start a server: diff --git a/envs/snake_env/README.md b/envs/snake_env/README.md index d08c30e20..aae66d0a0 100644 --- a/envs/snake_env/README.md +++ b/envs/snake_env/README.md @@ -19,6 +19,8 @@ A multi-agent snake game environment for OpenEnv, based on [marlenv](https://git *The `/web` playground after a few steps: the snake, the fruit and one button per move.* +Try it on the [`openenv/snake_env`](https://huggingface.co/spaces/openenv/snake_env) Space, or connect a client to `https://openenv-snake-env.hf.space`. + ## Overview The Snake environment wraps the marlenv Snake-v1 environment to provide a clean OpenEnv-compatible interface. Multiple snakes can battle on a fixed size grid map, but this implementation focuses on single-agent gameplay. diff --git a/envs/sumo_rl_env/README.md b/envs/sumo_rl_env/README.md index 3cc7fd432..cb279b020 100644 --- a/envs/sumo_rl_env/README.md +++ b/envs/sumo_rl_env/README.md @@ -15,6 +15,8 @@ tags: Integration of traffic signal control with the OpenEnv framework via SUMO (Simulation of Urban MObility) and SUMO-RL. +Try it on the [`openenv/sumo_rl_env`](https://huggingface.co/spaces/openenv/sumo_rl_env) Space, or connect a client to `https://openenv-sumo-rl-env.hf.space`. + ## Overview This environment enables reinforcement learning for **traffic signal control** using SUMO, a microscopic traffic simulation package. Train RL agents to optimize traffic light timing and minimize vehicle delays. diff --git a/envs/tbench2_env/README.md b/envs/tbench2_env/README.md index 03716b5db..c5ef7452c 100644 --- a/envs/tbench2_env/README.md +++ b/envs/tbench2_env/README.md @@ -22,6 +22,8 @@ OpenEnv wrapper for [Terminal-Bench 2](https://github.com/laude-institute/termin | **Local** | Runs commands in the server process (no Docker) | Hugging Face Spaces, environments without Docker access | | **Docker** | Runs each task in its own container | Full TB2.0 fidelity with custom task images | +Try it on the [`openenv/tbench2`](https://huggingface.co/spaces/openenv/tbench2) Space, or connect a client to `https://openenv-tbench2.hf.space`. + ## Quick Start ```python