Skip to content

Latest commit

 

History

History
55 lines (40 loc) · 1.89 KB

model-output.md

File metadata and controls

55 lines (40 loc) · 1.89 KB

Choose what the model sees

The Python SDK keeps full station data and supplies a concise model view by default. Observations include measured joints, tool pose, gripper state, faults, interlocks and configured camera images. Motion results retain outcomes, achieved poses and limits. Internal transport IDs and repeated history are omitted. Model values use millimetres and degrees; raw SDK data remains in SI.

The default robot tools are action_sequence, preview_pose, observe and stop. Tool selection and output formatting are independent. The Python example filters sequences to Cartesian movement and gripper actions and adds its own finish tool.

Full output

Starting with your existing robot, pass a different view to the provider adapter:

from hebbian_rua import full_formatter
from hebbian_rua.client.llm.openai import OpenAIRobotTools

model_tools = OpenAIRobotTools(robot.with_output_formatter(full_formatter))

Full output retains original fields, units and precision. Application code can always inspect result.data independently of the model view.

Customize presentation

Compose ordinary transforms before rendering:

from hebbian_rua import (
    ModelContent,
    RobotOutput,
    render_output,
    select_output_cameras,
    summarize_output,
)


def task_formatter(output: RobotOutput) -> ModelContent:
    output = summarize_output(output)
    output = select_output_cameras(output, {"table"})
    return render_output(output)


model_tools = robot.with_output_formatter(task_formatter)

The new view shares the station and lease and leaves the original unchanged. Transforms preserve SI units; rendering converts once. Camera selection changes model images without changing capture or required-camera checks. Formatting cannot change execution flags. Formatter exceptions propagate after execution and do not retry a robot command.