Annotating here some breaking changes to do in a single major version bump
[ ] Agent trace logger should not be passed to the planner, planner logs are useless for debugging, more useful is know events of planning and running a plan, and what the state changes are in between, that information can also be used for debugging in a separate environment.
[ ] Agent and sensor observables return a full copy of the state, however this requires a structuredClone of the state to avoid clients modifying the internal state of the agent, but moreover, the results on the agent observable can be very messy, as two sensors may report at almost exactly at the same time, resulting in the cloned object flipping values. TODO: link example later
Instead the agent observable should return changes as they happen in the form of operations (op, path,value). Users may act on the changes, we can export a utility to patch a state object using the path.
This will be good for testing as well, as the list of changes should equal the final state reported by the agent.
[ ] not really breaking, but it would be good to interrupt planning if it takes over some length of time, we already have maxSearchDepth, but preventing the agent from getting stuck in the planning stage seems worth it
[ ] Improve agent logging, perhaps using open telemetry
Annotating here some breaking changes to do in a single major version bump
(op, path,value)
. Users may act on the changes, we can export a utility to patch a state object using the path.