dropbear 0.1.0a2

Cloud robot policy inference — one function call.

Dropbear

Cloud robot policy inference — one function call.

0.1.0a2 is a pre-alpha release. Pin the exact version while the public API is still evolving.

Install

Add Dropbear to a Python 3.11–3.14 project:

uv add "dropbear==0.1.0a2"

Install the CLI as a standalone tool:

uv tool install "dropbear==0.1.0a2"
dropbear login

The base package contains cloud inference, transport, model contracts, and robot-neutral observation helpers. Install a hardware or simulator stack only when you need it:

uv add "dropbear[so101]==0.1.0a2"
uv add "dropbear[sim]==0.1.0a2"

The base SDK supports Python 3.11 through 3.14. The so101 extra currently supports Python 3.12 and 3.13 because its pinned LeRobot dependency requires Python 3.12 and does not yet provide a wheel-compatible Python 3.14 dependency stack.

The sim extra is currently Linux/WSL2 only because of upstream LIBERO/robosuite limitations.

Cloud policy inference

import dropbear

policy = dropbear.connect()  # defaults to molmoact2-so101
policy.run(
    instruction="pick up the cube",
    observe=observe,
    act=act,
    max_actions=300,
)

dropbear.connect() is the canonical SDK entrypoint. connect_so101() is a deprecated compatibility path for the legacy physical SO-101 safety loop and will be removed after that behavior is folded into the generic policy surface. New cloud-policy work should use dropbear.connect(). There is no first-class public connect_libero() or connect_franka() entrypoint; use dropbear.connect(model="molmoact2-libero") or select another checkpoint with model=.

Use a context manager so every cloud session closes deterministically:

import dropbear

with dropbear.connect(model="molmoact2-libero") as policy:
    result = policy.run(
        instruction="put the mug on the plate",
        observe=observe,
        act=act,
        max_actions=220,
        strategy=dropbear.RunStrategy.libero_default(),
    )

region="nearest" is the default. region="available" may fall back to another supported region when the nearest region has no capacity. Passing an AWS region pins the session there.

transport="auto" tries QUIC first and can use the hosted relay; transport="quic" requires QUIC, and transport="relay" uses the relay directly.

MolmoAct2-DROID on Franka

MolmoAct2-DROID consumes an exterior RGB view and wrist RGB view. A second exterior view is optional; when omitted, Dropbear reuses the first exterior frame for the checkpoint's second exterior slot.

Robot state is seven Franka joint positions in radians followed by a gripper value in [0, 1].

import numpy as np
import dropbear

exterior_rgb = np.zeros((480, 640, 3), dtype=np.uint8)
wrist_rgb = np.zeros((480, 640, 3), dtype=np.uint8)

observation = dropbear.franka.observe(
    exterior_frame=exterior_rgb,
    wrist_frame=wrist_rgb,
    joint_positions=[0.0] * 7,
    gripper=0.5,  # 0=open, 1=closed
)

with dropbear.connect(model="molmoact2-droid") as policy:
    result = policy.predict(
        observation,
        instruction="pick up the green block",
    )

assert len(result.actions) == 15
assert all(len(action) == 8 for action in result.actions)

The checkpoint returns a 15-step chunk at 15 Hz. Each action is an absolute target [joint_0, ..., joint_6, gripper]; joints are radians and the gripper is in [0, 1].

predict() does not actuate hardware or provide a Franka safety controller. Validate joint, velocity, acceleration, workspace, collision, and gripper limits before sending any target to a robot.

MolmoAct2-BimanualYAM

This surface was introduced in 0.1.0a2; 0.1.0a1 does not include it.

The BimanualYAM checkpoint consumes three RGB views in exact top, left, right order. State and absolute actions are 14 values: [left_joint_0..5, left_gripper, right_joint_0..5, right_gripper], with joints in radians and each gripper in [0, 1].

observation = dropbear.yam.observe(
    top_frame=top_rgb,
    left_frame=left_rgb,
    right_frame=right_rgb,
    left_joint_positions=left_joint_radians,
    left_gripper=left_gripper,
    right_joint_positions=right_joint_radians,
    right_gripper=right_gripper,
)

with dropbear.connect(model="molmoact2-bimanual-yam") as policy:
    result = policy.predict(observation, instruction="fold the towel")

The checkpoint returns 30 actions at 30 Hz. predict() performs inference only; the caller remains responsible for synchronized two-arm validation, soft limits, collision handling, watchdogs, and physical e-stop procedures. See the complete single-chunk example in examples/yam_inference.py.

The model is staged behind empty Terraform region gates and a coming_soon catalog entry. This SDK contract does not make it allocatable as a live session.

Manual action loop

For a caller-owned control loop, pass the task instruction to policy.next_action(...). Dropbear owns inference, refill, action buffering, calibration, and RTC prefix context; the caller owns sensing, actuation, and loop cadence.

import time

dt = 1.0 / policy.action_hz
while running:
    tick = time.perf_counter()
    action = policy.next_action(
        observe(),
        instruction="put the mug on the plate",
    )
    robot.execute(action)
    time.sleep(max(0.0, dt - (time.perf_counter() - tick)))

CLI

Sign in once. Credentials are stored in ~/.dropbear/config.toml.

dropbear login
dropbear status

For non-interactive setup, pass an API key or use bare --api-key to paste it into a hidden prompt:

dropbear login --api-key dropbear_sk_...
dropbear login --api-key

Run robot-neutral setup checks:

dropbear doctor

SO-101 and simulation checks are explicit:

dropbear doctor so101
dropbear doctor sim

Install shell completion with:

dropbear --install-completion

Useful session commands:

dropbear sessions list
dropbear sessions stop <session-id>
dropbear sessions stop --all

Documentation: https://dropbear.dreamscalelabs.com/docs