The app runs AMD using specified parameters against the Asterisk server it is connected to.
The samples need to be in a place that is reachable by Asterisk, in the path defined by config.amd_tuner.samples_dir.
The runner script in the script/ folder provides an interface that allows for parallel execution, dumping of results to a CSV file and more fine-grained control.
$ bundle exec script/runner --help
Usage: runner [file_id|ALL] [options]
-v, --[no-]verbose Run verbosely
-s, --[no-]summary Print summary
-c, --[no-]csv Write results to CSV
--num_threads N num_threads - Default: 3
--initial_silence N initial_silence - Default: 2500
--greeting N greeting - Default: 1500
--after_greeting_silence N after_greeting_silence - Default: 800
--total_analysis_time N total_analysis_time - Default: 5000
--min_word_length N min_word_length - Default: 100
--between_words_silence N between_words_silence - Default: 50
--maximum_number_of_words N maximum_number_of_words - Default: 3
--silence_threshold N silence_threshold - Default: 256
-h, --help Display this screen
Results are written to logs/amd_tuning.csv.
3 is the optimal number of threads for low-powered VMs, always pay attention to the increasing average detection time if looking to raise the number.
Reachable at http://AHN_IP:8080 (configured by the usual Virginia configuration), it is slow for running all the files but can provide a nice feedback.