Osprey's data archiver interface includes a downsampling option, which defaults to sampling every second:
https://github.com/als-apg/osprey/blob/main/src/osprey/connectors/archiver/mock_archiver_connector.py#L76
My experience is that the LLM does not change this default (even when I request data over a one year span, for which a 1 second sampling is a bit overkill!), unless if it is explicitly asked to do so.
However, some facility data maybe we measured at a slower rate and/or may have intermittent data e.g. as in the example here:
In some of these case, it may not even be necessary or sensible to apply downsampling.
I wonder if there could be a better default in Osprey, or if there are ways for the LLM to make a more sensible change of sampling - depending on the time range that it is being asked to retrieve.
Osprey's data archiver interface includes a downsampling option, which defaults to sampling every second:
https://github.com/als-apg/osprey/blob/main/src/osprey/connectors/archiver/mock_archiver_connector.py#L76
My experience is that the LLM does not change this default (even when I request data over a one year span, for which a 1 second sampling is a bit overkill!), unless if it is explicitly asked to do so.
However, some facility data maybe we measured at a slower rate and/or may have intermittent data e.g. as in the example here:
In some of these case, it may not even be necessary or sensible to apply downsampling.
I wonder if there could be a better default in Osprey, or if there are ways for the LLM to make a more sensible change of sampling - depending on the time range that it is being asked to retrieve.