Dressage
Scalable RL for Any Agent and Sandbox.
File Explorer
- docker.yml
- bench.png
- claw_eval_4metrics.png
- failed_32k_reward_segments.png
- success_64k_reward_segments.png
- context-window-segment-expansion.png
- glm-master-memory-prediction.png
- peak-memory-summary.png
- scaled-master-data-plane-memory-overview.png
- transfer-queue-data-path-en.png
- transfer-queue-data-path-zh.png
- alfworld_async.png
- alfworld_sync.png
- artchitecture.png
- dressage_logo.png
- hotpotqa_async.png
- hotpotqa_sync.png
- mopd_open_dynamics_8panel.png
- swegym-training-metrics.png
- version_span_results.png
- version_window_results.png
- __init__.py
- base.py
- claude_code.py
- codex.py
- factory.py
- openclaw.py
- opencode.py
- __init__.py
- health.py
- rollout.py
- sessions.py
- __init__.py
- command.py
- errors.py
- hashing.py
- models.py
- monitoring.py
- server.py
- __init__.py
- rollout_llm_proxy.py
- __init__.py
- paths.py
- __init__.py
- session_store.py
- __init__.py
- app.py
- config.py
- main.py
- pyproject.toml
- build.sh
- Dockerfile
- image_tag.sh
- README.md
- run.sh
- blackbox-dressage-claw-experiment-en.md
- blackbox-dressage-claw-experiment-zh.md
- blackbox-server.md
- blackbox-swegym-claude-code-experiment-en.md
- blackbox-swegym-claude-code-experiment-zh.md
- harbor.md
- mopd.md
- paddock.md
- proxy.md
- quickstart.md
- recipes.md
- rollout.md
- sandbox.md
- staleness.md
- swegym-claude-code-quickstart-en.md
- swegym-claude-code-quickstart-zh.md
- tool_call_hooks_proposal.md
- training.md
- trajectory-storage-refactor-en.md
- trajectory-storage-refactor-zh.md
- whitebox-agent-quickstart.md
- __init__.py
- config.py
- __init__.py
- artifacts.py
- compat.py
- config.py
- data_source.py
- environment.py
- gateway.py
- plugin.py
- rollout.py
- __init__.py
- __init__.py
- command.py
- defaults.py
- http_retry.py
- state.py
- utils.py
- __init__.py
- client.py
- execute_hooks.py
- failures.py
- paddock.py
- __init__.py
- paddock.py
- tools.py
- __init__.py
- factory.py
- interface.py
- lifecycle.py
- qwen3_5_mask_only_chat_template.jinja
- __init__.py
- prompt_assistant_mask.py
- qwen3_5_fixed.jinja
- __init__.py
- template_utils.py
- tito_tokenizer.py
- __init__.py
- generation_controller.py
- proxy_client.py
- reasoning_parser.py
- routed_experts.py
- server.py
- session_manager.py
- sglang_client.py
- tool_call_hooks.py
- tool_call_ids.py
- tool_call_parser.py
- trajectory_store.py
- __init__.py
- agent_whitebox.py
- reward.py
- tools.py
- __init__.py
- dispatch.py
- reward.py
- __init__.py
- agent_whitebox.py
- reward.py
- tools.py
- __init__.py
- evaluator.py
- integrity.py
- prepare_claude_code_sandbox_artifacts.sh
- reward.py
- __init__.py
- README.md
- __init__.py
- custom_rm.py
- helpers.py
- registry.py
- __init__.py
- samples.py
- writer.py
- __init__.py
- blackbox_dispatch.py
- blackbox_dispatch_swegym.py
- runtime.py
- whitebox_agent.py
- __init__.py
- config.py
- scheduler.py
- store.py
- __init__.py
- convert_samples.py
- data_source.py
- fully_async_rollout.py
- log_rollout.py
- mopd.py
- multi_segment.py
- partial_async_rollout.py
- staleness.py
- sync_rollout.py
- __init__.py
- manager.py
- provider.py
- runner.py
- slot.py
- supervisor.py
- __init__.py
- __init__.py
- provider.py
- __init__.py
- __init__.py
- blackbox_cluster_status.py
- local_bwrap_status.py
- start_blackbox_cluster.py
- start_local_bwrap.py
- stop_blackbox_cluster.py
- stop_local_bwrap.py
- __init__.py
- factory.py
- provider.py
- types.py
- __init__.py
- log_helpers.py
- mopd_launch_config.py
- mopd_loss.py
- mopd_megatron_actor.py
- mopd_train.py
- reward_post_process.py
- tq_hydration.py
- tq_megatron_actor.py
- tq_train.py
- tq_train_async.py
- train_async_with_rollout_pause.py
- __init__.py
- fields.py
- layouts.py
- runtime.py
- store.py
- __init__.py
- .gitignore
- prepare_alfworld.py
- build_e2b_blackbox_template.py
- prepare_dressage_claw_data.py
- .gitignore
- build_corpus.py
- build_index.py
- prepare_hotpotqa.py
- mopd_alfworld_hotpotqa.example.json
- prepare_swegym_data.py
- prepare_swegym_e2b.py
- .gitignore
- dressage_dapo_prompts.jsonl
- dressage_dapo_prompts_e2b.jsonl
- prepare_dataset.py
- rollout-bwrap.yaml
- rollout-native-local.yaml
- rollout-native-remote.yaml
- training-bwrap.yaml
- training-native-local.yaml
- training-native-remote.yaml
- tau3-bench-e2b.yaml
- terminal-bench-2-e2b.yaml
- run_harbor_rollout_qwen3.5_4b.sh
- run_harbor_training_qwen3.5_4b.sh
- dressage_env_defaults.sh
- dressage_staleness.yaml
- dressage_transfer_queue.yaml
- run_alfworld_whitebox_agent_qwen3.5_4b.sh
- run_alfworld_whitebox_agent_qwen3.5_4b_async.sh
- run_alfworld_whitebox_agent_qwen3.5_4b_partial_rollout_async.sh
- run_blackbox_qwen3.5_35b_a3b_sync_local.sh
- run_blackbox_qwen3.5_35b_a3b_sync_remote.sh
- run_blackbox_qwen3.5_4b_async_local.sh
- run_blackbox_qwen3.5_4b_async_remote.sh
- run_blackbox_qwen3.5_4b_partial_rollout_async_local.sh
- run_blackbox_qwen3.5_4b_partial_rollout_async_remote.sh
- run_dressage_claw_qwen3.6_35b_a3b_sync_4_node.sh
- run_dressage_swegym_qwen3.5_4b_claude_code_sync_4_node.sh
- run_hotpotqa_whitebox_agent_qwen3.5_4b.sh
- run_hotpotqa_whitebox_agent_qwen3.5_4b_async.sh
- run_mopd_qwen3.5_sync.sh
- tool_call_hooks_example.py
- __init__.py
- test_async_turns.py
- test_claude_code_adapter.py
- test_codex_adapter.py
- test_command.py
- test_openclaw_adapter.py
- test_opencode_adapter.py
- test_rollout_llm_proxy.py
- test_server.py
- conftest.py
- test_contracts.py
- test_data_path.py
- test_environment.py
- test_execution.py
- test_gated_e2e.py
- __init__.py
- conftest.py
- test_async_rollout_worker_lifecycle.py
- test_blackbox_dispatch.py
- test_blackbox_dispatch_multi_segment.py
- test_blackbox_node_supervisor.py
- test_config.py
- test_convert_samples_multi_segment.py
- test_custom_rm.py
- test_data_source.py
- test_docker_release.py
- test_dressage_claw_dispatch.py
- test_fully_async_rollout.py
- test_log_rollout_data.py
- test_mopd.py
- test_mopd_loss.py
- test_mopd_metrics.py
- test_multi_segment.py
- test_new_paddock_layers.py
- test_paddock_blackbox_defaults.py
- test_paddock_multi_segment.py
- test_partial_async_rollout.py
- test_prewarm_provider_integration.py
- test_prewarm_scheduler.py
- test_prewarm_store.py
- test_proxy.py
- test_proxy_client.py
- test_ray_blackbox_scheduler.py
- test_resume_readiness_simple.py
- test_reward_post_process_multi_segment.py
- test_reward_registry.py
- test_rollout_helper_modules.py
- test_routed_experts.py
- test_sandbox_provider_layer.py
- test_sandbox_runner.py
- test_session_manager_cleanup.py
- test_slime_compat.py
- test_slot_archives.py
- test_staleness.py
- test_tito.py
- test_tool_call_hooks.py
- test_tq_sample_layout.py
- test_tq_training_hydration.py
- test_training_arguments.py
- test_transport_core.py
- .dockerignore
- .gitignore
- .gitmodules
- LICENSE
- NOTICE
- pyproject.toml
- README.md
- slime
# Use via CDN
jsDelivrjsDelivr serves any public GitHub repository as a CDN with zero setup. Pick a version and a file to get a ready-to-paste link and snippet.
Command Glossary
Commands referenced in this DOCs, explained below.
codex
View Details ▼
codex
Natural language code assistant for the terminal, powered by OpenAI.
Reads and edits files in your current directory to fulfill requests.
codex
Start an interactive Codex session in the current directory:
codex "{{prompt}}"
Run a single Codex command using a prompt:
codex --sandbox workspace-write "{{prompt}}"
Run a prompt allowing Codex to edit files in the workspace:
docker pull
View Details ▼
docker pull
This command is an alias of `docker image pull`.
tldr docker image pull
View documentation for the original command:
docker run
View Details ▼
docker run
This command is an alias of `docker container run`.
tldr docker container run
View documentation for the original command:
docker
View Details ▼
docker
Manage Docker containers and images.
Some subcommands such as `container` and `image` have their own usage documentation.
docker {{[ps|container ls]}} {{[-a|--all]}}
List all Docker containers (running and stopped):
docker {{[run|container run]}} --name {{container_name}} {{image}}
Start a container from an image, with a custom name:
docker container {{start|stop}} {{container_name}}
Start or stop an existing container:
git checkout
View Details ▼
git checkout
Checkout a branch or paths to the working tree.
git checkout -b {{branch_name}}
Create and switch to a new branch:
git checkout -b {{branch_name}} {{reference}}
Create and switch to a new branch based on a specific reference (branch, remote/branch, tag are examples of valid references):
git checkout {{branch_name}}
Switch to an existing local branch:
git commit
View Details ▼
git commit
Commit files to the repository.
git commit
Open an editor to write a message and commit staged files to the repository:
git commit {{[-m|--message]}} "{{message}}"
Commit staged files to the repository with the specified message:
git commit {{[-F|--file]}} {{path/to/commit_message_file}}
Commit staged files with a message read from a file:
git push
View Details ▼
git push
Push commits to a remote repository.
git push
Send local changes in the current branch to its default remote counterpart:
git push {{remote_name}} {{local_branch}}
Send changes from a specific local branch to its remote counterpart:
git push {{[-u|--set-upstream]}} {{remote_name}} {{local_branch}}
Send changes from a specific local branch to its remote counterpart, and set the remote one as the default push/pull target of the local one:
git submodule
View Details ▼
git submodule
Inspect, update, and manage submodules.
git submodule
View existing submodules and the checked-out commit for each one:
git submodule update --init --recursive
Install a repository's submodules (listed in `.gitmodules`):
git submodule add {{repository_url}}
Add a Git repository as a submodule of the current one:
openclaw
View Details ▼
openclaw
A personal AI assistant that you run on your own devices.
Some subcommands such as `onboard`, `agent`, `doctor`, etc. have their own usage documentation.
See also: `zeroclaw`.
openclaw onboard --install-daemon
Run the onboarding wizard to set up the gateway and channels:
openclaw gateway --allow-unconfigured --port {{18789}} --verbose
Start the gateway server:
openclaw message send {{[-t|--target]}} {{+1234567890}} {{[-m|--message]}} "{{Hello}}"
Send a message to a contact:
opencode
View Details ▼
opencode
An AI coding agent.
Some subcommands such as `auth`, `models`, `web`, etc. have their own usage documentation.
opencode
Start the interactive TUI:
opencode {{[-c|--continue]}}
Continue the most recent session:
opencode run "{{message}}"
Run opencode in non-interactive mode by passing a prompt directly:
pytest
View Details ▼
pytest
Run Python tests.
pytest {{path/to/test_file1.py path/to/test_file2.py ...}}
Run tests from specific files:
pytest -k {{expression}}
Run tests with names matching a specific [k]eyword expression:
pytest {{[-x|--exitfirst]}}
Exit as soon as a test fails or encounters an error:
