inspect-robots-xpolicylab 0.1.0__tar.gz

inspect_robots_xpolicylab-0.1.0/.gitignore

@@ -0,0 +1,221 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[codz]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#   Usually these files are written by a python script from a template
+#   before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py.cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+# Pipfile.lock
+
+# UV
+#   Similar to Pipfile.lock, it is generally recommended to include uv.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+# uv.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+# poetry.lock
+# poetry.toml
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#   pdm recommends including project-wide configuration in pdm.toml, but excluding .pdm-python.
+#   https://pdm-project.org/en/latest/usage/project/#working-with-version-control
+# pdm.lock
+# pdm.toml
+.pdm-python
+.pdm-build/
+
+# pixi
+#   Similar to Pipfile.lock, it is generally recommended to include pixi.lock in version control.
+# pixi.lock
+#   Pixi creates a virtual environment in the .pixi directory, just like venv module creates one
+#   in the .venv directory. It is recommended not to include this directory in version control.
+.pixi
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# Redis
+*.rdb
+*.aof
+*.pid
+
+# RabbitMQ
+mnesia/
+rabbitmq/
+rabbitmq-data/
+
+# ActiveMQ
+activemq-data/
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.envrc
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#   JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#   be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#   and can be added to the global gitignore or merged into this file.  For a more nuclear
+#   option (not recommended) you can uncomment the following to ignore the entire idea folder.
+# .idea/
+
+# Abstra
+#   Abstra is an AI-powered process automation framework.
+#   Ignore directories containing user credentials, local state, and settings.
+#   Learn more at https://abstra.io/docs
+.abstra/
+
+# Visual Studio Code
+#   Visual Studio Code specific template is maintained in a separate VisualStudioCode.gitignore
+#   that can be found at https://github.com/github/gitignore/blob/main/Global/VisualStudioCode.gitignore
+#   and can be added to the global gitignore or merged into this file. However, if you prefer,
+#   you could uncomment the following to ignore the entire vscode folder
+# .vscode/
+# Temporary file for partial code execution
+tempCodeRunnerFile.py
+
+# Ruff stuff:
+.ruff_cache/
+
+# PyPI configuration file
+.pypirc
+
+# Marimo
+marimo/_static/
+marimo/_lsp/
+__marimo__/
+
+# Streamlit
+.streamlit/secrets.toml
+
+# uv.lock is committed: CI installs with `uv sync --locked` for reproducible
+# builds. Run `uv lock` after changing dependencies in pyproject.toml.

inspect_robots_xpolicylab-0.1.0/PKG-INFO

@@ -0,0 +1,141 @@
+Metadata-Version: 2.4
+Name: inspect-robots-xpolicylab
+Version: 0.1.0
+Summary: XPolicyLab policy adapter for Inspect Robots — evaluate any XPolicyLab-served VLA policy (π0, GR00T, OpenVLA-OFT, RDT, ACT, ...) with Inspect Robots.
+Project-URL: Homepage, https://github.com/robocurve/inspect-robots/tree/main/plugins/inspect-robots-xpolicylab
+Project-URL: Repository, https://github.com/robocurve/inspect-robots
+Author: Inspect Robots contributors
+License-Expression: MIT
+Keywords: evaluation,inspect_robots,policy-server,robotics,vla,xpolicylab
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Science/Research
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Programming Language :: Python :: 3
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: Typing :: Typed
+Requires-Python: >=3.10
+Requires-Dist: inspect-robots>=0.2
+Requires-Dist: msgpack-numpy>=0.4.8
+Requires-Dist: msgpack>=1.0
+Requires-Dist: numpy>=1.24
+Requires-Dist: websockets>=12
+Provides-Extra: dev
+Requires-Dist: mypy>=1.11; extra == 'dev'
+Requires-Dist: pytest-cov>=5.0; extra == 'dev'
+Requires-Dist: pytest>=8.0; extra == 'dev'
+Requires-Dist: ruff>=0.6; extra == 'dev'
+Description-Content-Type: text/markdown
+
+# inspect-robots-xpolicylab
+
+Drive any [XPolicyLab](https://github.com/XPolicyLab/XPolicyLab)-served policy
+from [Inspect Robots](https://github.com/robocurve/inspect-robots).
+
+XPolicyLab unifies **40+ VLA and imitation-learning policies** (π0 / π0.5,
+GR00T N1.7, OpenVLA-OFT, RDT-1B, SmolVLA, ACT, Diffusion Policy, …) behind one
+policy-server contract: each policy runs in its own conda/uv environment and
+serves inference over a websocket protocol. This plugin is the Inspect Robots
+side of that wire: a `Policy` adapter that connects to a running XPolicyLab
+policy server, so the whole policy zoo becomes evaluable with any Inspect
+Robots embodiment and task.
+
+The adapter speaks XPolicyLab's msgpack-over-websocket protocol directly — it
+does **not** depend on the `xpolicylab` package (which is not on PyPI). Only
+the policy *server* needs an XPolicyLab checkout.
+
+## Install
+
+```bash
+pip install inspect-robots-xpolicylab
+```
+
+The `xpolicylab` policy then appears in `inspect-robots list policies`.
+
+## Quickstart
+
+**Terminal 1 — serve a policy from your XPolicyLab checkout** (its own
+environment, possibly another machine; see the
+[XPolicyLab README](https://github.com/XPolicyLab/XPolicyLab#readme) for
+per-policy install/checkpoint details):
+
+```bash
+cd XPolicyLab/policy/Pi_0
+bash setup_eval_policy_server.sh <bench_name> <task_name> <ckpt_name> \
+    <env_cfg_type> <action_type> <seed> <policy_gpu_id> <policy_env> 19000 0.0.0.0
+```
+
+**Terminal 2 — evaluate it with Inspect Robots:**
+
+```bash
+inspect-robots run --task my-task --policy xpolicylab --embodiment isaacsim \
+    -P url=ws://gpu-box:19000 -P cameras=cam_head:base_rgb
+```
+
+or programmatically:
+
+```python
+from inspect_robots import eval
+from inspect_robots_xpolicylab import XPolicyLabPolicy
+
+with XPolicyLabPolicy(url="ws://gpu-box:19000", cameras="cam_head:base_rgb") as policy:
+    log = eval("my-task", policy, "isaacsim")
+```
+
+Constructing the policy and reading `.info` never touch the network — the
+websocket connects on first `reset()`/`act()`, with retries while the policy
+server cold-starts (loading VLA weights can take minutes).
+
+## Configuration
+
+| Arg | Default | Meaning |
+| --- | --- | --- |
+| `url` | `ws://localhost:19000` | policy server websocket URL |
+| `action_type` | `joint` | `joint` → `*arm_joint_state` actions; `ee` → `*ee_pose` |
+| `arms` | `1` | `1` or `2` (dual-arm uses `left_*`/`right_*` keys) |
+| `arm_dim` / `ee_dim` | `7` / `1` | per-arm joint dims / gripper dims |
+| `cameras` | `cam_head:cam_head` | XPolicyLab camera slot → Inspect Robots camera name |
+| `state_map` | see below | XPolicyLab state key → Inspect Robots state key |
+| `required_state_keys` | joint: `joint_pos,gripper`; ee: none | state keys enforced by compatibility checking |
+| `action_keys` + `action_dim` | derived | explicit ordered action-dict keys for exotic setups |
+| `camera_height` / `camera_width` | unset | declare camera resolution for compatibility checks |
+| `control_hz` | unset | chunk playback rate; also sent as `additional_info.frequency` |
+| `name` | `xpolicylab` | policy name in logs (tag runs e.g. `xpolicylab:pi0`) |
+| `connect_timeout_s` / `request_timeout_s` | `30` / `120` | client timeouts |
+| `connect_attempts` / `connect_retry_delay_s` | `10` / `5` | cold-start retry budget |
+
+Mapping-valued args accept a compact string form for the CLI:
+`-P cameras=cam_head:base_rgb,cam_wrist:wrist_rgb`.
+
+Default `state_map` (all XPolicyLab state fields are optional; mapped keys
+missing from an observation are simply not sent):
+
+| XPolicyLab key | Inspect Robots key |
+| --- | --- |
+| `arm_joint_state` | `joint_pos` |
+| `ee_joint_state` | `gripper` |
+| `ee_pose` | `eef_pose` |
+
+## Action mapping
+
+Each `infer` reply is an action chunk — one dict per future control step.
+Dicts flatten to vectors in a fixed order: for each arm (left then right), the
+arm key then the end-effector key. The default single-arm joint profile is
+`concat(arm_joint_state, ee_joint_state)` → dim 8, matching the
+`inspect-robots-isaacsim` default Franka profile.
+
+## Lifecycle notes
+
+- `reset(scene)` ends the previous trial server-side (`trial_end`) and starts
+  a fresh `trial_id`; `close()` ends any open trial and says goodbye.
+- `eval()` closes embodiments it resolves, **not** policies — use `with`
+  / `close()` in programmatic runs. A best-effort `atexit` hook covers
+  registry-resolved CLI runs.
+- If the socket drops mid-eval, the trial in flight errors and the next
+  `reset()`/`act()` reconnects (replaying the `hello` handshake).
+
+## Protocol compatibility
+
+Validated against XPolicyLab commit
+`fe71eb54675cef495fea817a637386a4f4529153`. The protocol carries no version
+field; if upstream changes the wire format, `_protocol.py` is the single small
+module to diff and update.

inspect_robots_xpolicylab-0.1.0/README.md

@@ -0,0 +1,113 @@
+# inspect-robots-xpolicylab
+
+Drive any [XPolicyLab](https://github.com/XPolicyLab/XPolicyLab)-served policy
+from [Inspect Robots](https://github.com/robocurve/inspect-robots).
+
+XPolicyLab unifies **40+ VLA and imitation-learning policies** (π0 / π0.5,
+GR00T N1.7, OpenVLA-OFT, RDT-1B, SmolVLA, ACT, Diffusion Policy, …) behind one
+policy-server contract: each policy runs in its own conda/uv environment and
+serves inference over a websocket protocol. This plugin is the Inspect Robots
+side of that wire: a `Policy` adapter that connects to a running XPolicyLab
+policy server, so the whole policy zoo becomes evaluable with any Inspect
+Robots embodiment and task.
+
+The adapter speaks XPolicyLab's msgpack-over-websocket protocol directly — it
+does **not** depend on the `xpolicylab` package (which is not on PyPI). Only
+the policy *server* needs an XPolicyLab checkout.
+
+## Install
+
+```bash
+pip install inspect-robots-xpolicylab
+```
+
+The `xpolicylab` policy then appears in `inspect-robots list policies`.
+
+## Quickstart
+
+**Terminal 1 — serve a policy from your XPolicyLab checkout** (its own
+environment, possibly another machine; see the
+[XPolicyLab README](https://github.com/XPolicyLab/XPolicyLab#readme) for
+per-policy install/checkpoint details):
+
+```bash
+cd XPolicyLab/policy/Pi_0
+bash setup_eval_policy_server.sh <bench_name> <task_name> <ckpt_name> \
+    <env_cfg_type> <action_type> <seed> <policy_gpu_id> <policy_env> 19000 0.0.0.0
+```
+
+**Terminal 2 — evaluate it with Inspect Robots:**
+
+```bash
+inspect-robots run --task my-task --policy xpolicylab --embodiment isaacsim \
+    -P url=ws://gpu-box:19000 -P cameras=cam_head:base_rgb
+```
+
+or programmatically:
+
+```python
+from inspect_robots import eval
+from inspect_robots_xpolicylab import XPolicyLabPolicy
+
+with XPolicyLabPolicy(url="ws://gpu-box:19000", cameras="cam_head:base_rgb") as policy:
+    log = eval("my-task", policy, "isaacsim")
+```
+
+Constructing the policy and reading `.info` never touch the network — the
+websocket connects on first `reset()`/`act()`, with retries while the policy
+server cold-starts (loading VLA weights can take minutes).
+
+## Configuration
+
+| Arg | Default | Meaning |
+| --- | --- | --- |
+| `url` | `ws://localhost:19000` | policy server websocket URL |
+| `action_type` | `joint` | `joint` → `*arm_joint_state` actions; `ee` → `*ee_pose` |
+| `arms` | `1` | `1` or `2` (dual-arm uses `left_*`/`right_*` keys) |
+| `arm_dim` / `ee_dim` | `7` / `1` | per-arm joint dims / gripper dims |
+| `cameras` | `cam_head:cam_head` | XPolicyLab camera slot → Inspect Robots camera name |
+| `state_map` | see below | XPolicyLab state key → Inspect Robots state key |
+| `required_state_keys` | joint: `joint_pos,gripper`; ee: none | state keys enforced by compatibility checking |
+| `action_keys` + `action_dim` | derived | explicit ordered action-dict keys for exotic setups |
+| `camera_height` / `camera_width` | unset | declare camera resolution for compatibility checks |
+| `control_hz` | unset | chunk playback rate; also sent as `additional_info.frequency` |
+| `name` | `xpolicylab` | policy name in logs (tag runs e.g. `xpolicylab:pi0`) |
+| `connect_timeout_s` / `request_timeout_s` | `30` / `120` | client timeouts |
+| `connect_attempts` / `connect_retry_delay_s` | `10` / `5` | cold-start retry budget |
+
+Mapping-valued args accept a compact string form for the CLI:
+`-P cameras=cam_head:base_rgb,cam_wrist:wrist_rgb`.
+
+Default `state_map` (all XPolicyLab state fields are optional; mapped keys
+missing from an observation are simply not sent):
+
+| XPolicyLab key | Inspect Robots key |
+| --- | --- |
+| `arm_joint_state` | `joint_pos` |
+| `ee_joint_state` | `gripper` |
+| `ee_pose` | `eef_pose` |
+
+## Action mapping
+
+Each `infer` reply is an action chunk — one dict per future control step.
+Dicts flatten to vectors in a fixed order: for each arm (left then right), the
+arm key then the end-effector key. The default single-arm joint profile is
+`concat(arm_joint_state, ee_joint_state)` → dim 8, matching the
+`inspect-robots-isaacsim` default Franka profile.
+
+## Lifecycle notes
+
+- `reset(scene)` ends the previous trial server-side (`trial_end`) and starts
+  a fresh `trial_id`; `close()` ends any open trial and says goodbye.
+- `eval()` closes embodiments it resolves, **not** policies — use `with`
+  / `close()` in programmatic runs. A best-effort `atexit` hook covers
+  registry-resolved CLI runs.
+- If the socket drops mid-eval, the trial in flight errors and the next
+  `reset()`/`act()` reconnects (replaying the `hello` handshake).
+
+## Protocol compatibility
+
+Validated against XPolicyLab commit
+`fe71eb54675cef495fea817a637386a4f4529153`. The protocol carries no version
+field; if upstream changes the wire format, `_protocol.py` is the single small
+module to diff and update.

inspect_robots_xpolicylab-0.1.0/pyproject.toml

@@ -0,0 +1,100 @@
+[build-system]
+requires = ["hatchling>=1.18", "hatch-fancy-pypi-readme>=24.1"]
+build-backend = "hatchling.build"
+
+[project]
+name = "inspect-robots-xpolicylab"
+version = "0.1.0"
+description = "XPolicyLab policy adapter for Inspect Robots — evaluate any XPolicyLab-served VLA policy (π0, GR00T, OpenVLA-OFT, RDT, ACT, ...) with Inspect Robots."
+dynamic = ["readme"]
+requires-python = ">=3.10"
+license = "MIT"
+authors = [{ name = "Inspect Robots contributors" }]
+keywords = ["robotics", "xpolicylab", "vla", "policy-server", "inspect_robots", "evaluation"]
+classifiers = [
+    "Development Status :: 3 - Alpha",
+    "Intended Audience :: Science/Research",
+    "License :: OSI Approved :: MIT License",
+    "Programming Language :: Python :: 3",
+    "Topic :: Scientific/Engineering :: Artificial Intelligence",
+    "Typing :: Typed",
+]
+# NOTE: `xpolicylab` itself is intentionally NOT a dependency. It is not published
+# on PyPI (git-install only) and ships collision-prone top-level packages
+# (`client_server`, `utils`). This adapter speaks XPolicyLab's small websocket
+# protocol directly; the policy server always runs from the user's XPolicyLab
+# checkout in its own environment — that separation is XPolicyLab's core design.
+#
+# websockets floor: the sync client this adapter uses exists since v11; Isaac Sim
+# environments (which the flagship isaacsim pairing shares a venv with) ship v12.
+dependencies = [
+    "inspect-robots>=0.2",
+    "numpy>=1.24",
+    "websockets>=12",
+    "msgpack>=1.0",
+    "msgpack-numpy>=0.4.8",
+]
+
+[project.optional-dependencies]
+dev = ["pytest>=8.0", "pytest-cov>=5.0", "mypy>=1.11", "ruff>=0.6"]
+
+# Entry-point discovery: an installed inspect-robots-xpolicylab appears in
+# `inspect-robots list policies` without being imported first. The registry calls
+# this factory by the name `xpolicylab`.
+[project.entry-points."inspect_robots.policies"]
+xpolicylab = "inspect_robots_xpolicylab:xpolicylab_policy"
+
+[project.urls]
+Homepage = "https://github.com/robocurve/inspect-robots/tree/main/plugins/inspect-robots-xpolicylab"
+Repository = "https://github.com/robocurve/inspect-robots"
+
+# In the Inspect Robots monorepo this resolves the `inspect_robots` dependency to the
+# in-repo core package instead of PyPI. Harmless in a standalone checkout (uv
+# ignores a workspace source when there is no workspace root).
+[tool.uv.sources]
+inspect-robots = { workspace = true }
+
+[tool.hatch.build.targets.wheel]
+packages = ["src/inspect_robots_xpolicylab"]
+
+[tool.hatch.build.targets.sdist]
+include = ["src/inspect_robots_xpolicylab", "tests", "README.md"]
+
+[tool.ruff]
+line-length = 100
+
+[tool.mypy]
+strict = true
+
+# msgpack / msgpack-numpy ship no type stubs; the codec wraps them in one module.
+[[tool.mypy.overrides]]
+module = ["msgpack.*", "msgpack_numpy.*"]
+ignore_missing_imports = true
+
+# PyPI readme: identical to README.md except GitHub-only alert syntax
+# (e.g. `> [!NOTE]`) is rewritten to bold blockquotes, which PyPI renders.
+[tool.hatch.metadata.hooks.fancy-pypi-readme]
+content-type = "text/markdown"
+
+[[tool.hatch.metadata.hooks.fancy-pypi-readme.fragments]]
+path = "README.md"
+
+[[tool.hatch.metadata.hooks.fancy-pypi-readme.substitutions]]
+pattern = '(?m)^> \[!NOTE\][ \t]*$'
+replacement = '> **Note:**'
+
+[[tool.hatch.metadata.hooks.fancy-pypi-readme.substitutions]]
+pattern = '(?m)^> \[!TIP\][ \t]*$'
+replacement = '> **Tip:**'
+
+[[tool.hatch.metadata.hooks.fancy-pypi-readme.substitutions]]
+pattern = '(?m)^> \[!IMPORTANT\][ \t]*$'
+replacement = '> **Important:**'
+
+[[tool.hatch.metadata.hooks.fancy-pypi-readme.substitutions]]
+pattern = '(?m)^> \[!WARNING\][ \t]*$'
+replacement = '> **Warning:**'
+
+[[tool.hatch.metadata.hooks.fancy-pypi-readme.substitutions]]
+pattern = '(?m)^> \[!CAUTION\][ \t]*$'
+replacement = '> **Caution:**'

inspect_robots_xpolicylab-0.1.0/src/inspect_robots_xpolicylab/__init__.py

@@ -0,0 +1,42 @@
+"""inspect-robots-xpolicylab — drive XPolicyLab-served policies from Inspect Robots.
+
+`XPolicyLab <https://github.com/XPolicyLab/XPolicyLab>`_ unifies 40+ VLA and
+imitation-learning policies behind one websocket policy-server contract. Start
+any of its policy servers, then evaluate it like any other Inspect Robots
+policy::
+
+    inspect-robots list policies           # -> includes "xpolicylab"
+    inspect-robots run --task my-task --policy xpolicylab --embodiment isaacsim \
+        -P url=ws://gpu-box:19000 -P cameras=cam_head:base_rgb
+
+or programmatically::
+
+    from inspect_robots import eval
+    from inspect_robots_xpolicylab import XPolicyLabPolicy
+
+    with XPolicyLabPolicy(url="ws://gpu-box:19000") as policy:
+        eval("my-task", policy, "isaacsim")
+
+The policy is discovered via the ``inspect_robots.policies`` entry point, so it
+shows up without being imported first.
+"""
+
+from __future__ import annotations
+
+from typing import Any
+
+from inspect_robots_xpolicylab.policy import XPolicyLabPolicy
+
+__all__ = ["XPolicyLabPolicy", "xpolicylab_policy"]
+
+__version__ = "0.1.0"
+
+
+def xpolicylab_policy(**kwargs: Any) -> XPolicyLabPolicy:
+    """Factory the Inspect Robots registry calls (entry point ``xpolicylab``).
+
+    Accepts the same keyword arguments as :class:`XPolicyLabPolicy`; the CLI
+    forwards ``-P key=value`` pairs here (mapping-valued args accept the
+    compact ``"key:value,key:value"`` string form).
+    """
+    return XPolicyLabPolicy(**kwargs)