feat: MuJoCo simulation backend - AgentTool with 35 actions#85

Open

cagataycali wants to merge 90 commits intostrands-labs:mainfrom

cagataycali:feat/mujoco-backend

Commits on May 3, 2026

feat: MuJoCo simulation backend — AgentTool with 35 actions

cagataycali
authored and
strands-agent
committed
fix: address all review comments — ABC, thread-safety, injection, cleanup

cagataycali
authored and
strands-agent
committed
fix: resolve lint errors — import ordering, format, strict param

cagataycali
authored and
strands-agent
committed
fix: acquire _lock around MuJoCo data mutations + sanitize all XML names

cagataycali
authored and
strands-agent
committed
ci: add MuJoCo system deps (libosmesa6-dev + MUJOCO_GL=osmesa)

cagataycali
authored and
strands-agent
committed
feat: add [sim] extra with mujoco dependency

cagataycali
authored and
strands-agent
committed
fix: rename [sim] extra to [sim-mujoco] per review

strands-agent
authored and
strands-agent
committed
fix: rebase on simulation-foundation — SimulationBackend→SimEngine, update lazy imports

cagataycali
authored and
strands-agent
committed
fix: resolve all mypy errors — mixin overrides, Optional types, import stubs

cagataycali
authored and
strands-agent
committed
fix: properly fix mypy errors instead of blanket suppression

cagataycali
authored and
strands-agent
committed
fix: zero mypy suppressions — proper type declarations instead of disable_error_code

cagataycali
authored and
strands-agent
committed
feat(sim): use require_optional for imageio in policy_runner

cagataycali
authored and
strands-agent
committed
fix: address 8 review threads — deps, exports, init, headless, coupling, tests, XML parsing

cagataycali
authored and
strands-agent
committed
fix: replace Protocol annotation with direct TYPE_CHECKING stubs in PolicyRunnerMixin

strands-agent
authored and
strands-agent
committed
refactor(mujoco): migrate SimWorld private fields to _backend_state

cagataycali
authored and
strands-agent
committed
fix(mujoco): resolve 4 bugs — add_robot world model, eval_policy double step, run_policy lock, joint→ctrl mapping

cagataycali
authored and
strands-agent
committed
fix: sync sim_time/step_count in replay_episode and eval_policy, add 99 integration tests

cagataycali
authored and
strands-agent
committed
fix(mujoco): prevent C-level abort on headless without EGL/OSMesa

cagataycali
authored and
strands-agent
committed
fix(mujoco): resolve mesh path mismatch during robot injection

cagataycali
authored and
strands-agent
committed
fix(mujoco): forward observation_mapping/action_mapping through tool_spec dispatcher

cagataycali
authored and
strands-agent
committed
refactor(sim): extract backend-agnostic PolicyRunner

cagataycali
authored and
strands-agent
committed
fix(mujoco): make renderer cache thread-local to prevent CGL segfault

cagataycali
authored and
strands-agent
committed
test(mujoco): regression for renderer thread-safety (CGL segfault)

cagataycali
authored and
strands-agent
committed
fix(mujoco): reset mj_saveLastXML global state for all inject/eject paths

cagataycali
authored and
strands-agent
committed
feat(mujoco): support multiple same-config robots via XML namespacing

cagataycali
authored and
strands-agent
committed
fix(mujoco): physics + recording callsites for namespaced entities

cagataycali
authored and
strands-agent
committed
refactor(sim): tighten get_observation schema, drop camera_name param

cagataycali
authored and
strands-agent
committed
refactor(sim): consolidate run_policy video params + XML injection fuzzer

cagataycali
authored and
strands-agent
committed
fix: address 10 blocking review issues from @yinsong1986

cagataycali
authored and
strands-agent
committed
fix: extend thread-safety locks, correct apply_force docstring, add type guard

cagataycali
authored and
strands-agent
committed
test: add camera roundtrip + multi-robot asset dir regression tests

cagataycali
authored and
strands-agent
committed
fix: address review feedback from @yinsong1986 (2026-05-01)

cagataycali
authored and
strands-agent
committed
style: drop date reference in reset comment (non-blocking nit)

strands-agent
authored and
strands-agent
committed
fix: block scene mutations while policy is running (concurrency guard)

strands-agent
authored and
strands-agent
committed
fix: add concurrency guards to move_object and remove_robot

cagataycali
authored and
strands-agent
committed
fix: narrow lock gap in set_body/geom_properties + add @requires_gl test guards

strands-agent
authored and
strands-agent
committed
fix: mock importlib in factory test so it passes when mujoco IS installed

strands-agent
authored and
strands-agent
committed
feat(sim): multi-camera capture, public DX polish, MP4 recording fix
cagataycali
committed
fix: update _stop_policy → stop_policy in regression tests
strands-agent
committed
test: mirror tests/ layout to strands_robots/ source tree
cagataycali
committed
fix(tests): add CI skipif guards to recording tests in test_rendering.py
cagataycali
committed
fix: resolve lint errors from test restructuring
cagataycali
committed
chore: strip emojis/dividers from logs+strings, fix leading-space artifacts
cagataycali
committed
feat(sim): prefix joint names per-robot in multi-robot recordings
cagataycali
committed
test: add coverage for error paths, module API, multi-robot recordings
cagataycali
committed
chore: apply ruff format/lint fixes
cagataycali
committed
fix(tests): use inline XML fixtures, avoid network-dependent so101 model
cagataycali
committed
fix(tests): use inline XML fixtures in test_recording_paths to avoid network-dependent so101 model
cagataycali
committed
fix(ci): add features=["all"] to hatch env, add importorskip guards
cagataycali
committed
fix(ci): install ffmpeg for torchcodec, add remaining importorskip guards
cagataycali
committed
fix(ci): install ffmpeg for torchcodec, handle video decode RuntimeError
cagataycali
committed
fix(tests): reduce policy duration and increase result timeout for CI
cagataycali
committed

Commits on May 4, 2026

Commits on May 5, 2026