Skip to content


The benchmarking tools allow measuring the performance of an agent service realised on the Open Autonomy framework.

How to use benchmarking tools

  1. Setup agent runtime environment in dev mode using autonomy deploy build KEYS --dev.

  2. Run agents for 1 period and wait until the round where the skill is configured to call the method. (You can also run agents for N periods if you want more data).

  3. Point the aggregation script to the directory containing the benchmark data. This will generate a benchmarks.html file containing benchmark stats in your current directory.

Example usage

Run the oracle/price_estimation service as explained here.

By default this will create a 4 agent runtime where you can wait until all 4 agents are at the end of the first period (you can wait for more periods if you want) and then you can stop the runtime. The data will be stored in the abci_build/persistent_data/benchmarks folder. You can use the following script to aggregate this data:

autonomy analyse benchmarks abci_build/persistent_data/benchmarks

By default the script will generate output for all periods but you can specify which period to generate output for. Similarly, block types aggregation is configurable as well.

Back to top