Hi @Anaisdg, the tutorial is interesting, but it is not clear how to use it in production and how to integrate it with Kapacitor. Which is the workflow?
You would write a UDF script that would include tensor flow and then you would configure kapactor to run it. Try taking a look at this. Please let me know if this clarifies things. Also, if you’re interested in making predictions, I suggest taking a look at the built in holt-winters predictor.
Can Kapacitor stream data to our Python code. If yes, what is the interface? Is it UDF only?
I am trying to run my ML forecast model for stream data. I want to use influxdb to store this stream data and my model should get the data from influx, predict and write back to influx. I know I can use python influxclient to do this, but I am looking for a way to run the model from the TICK and not as an independent python script. Can thoughts as to what I can use for this? UDF? Can UDF run keras?
Python isn’t currently supported server side with InfluxDB. However you can create a flux task and call your python script on an interval.
However, I recommend using the execd processor plugin above all for streaming forecasts.
Please take a look at the following resources for stream forecasting: