Automatic Dialer testbed

This repository contains a testbed for automatic dialing algorithms. It provides a simulation where algorithms can be experimented with, without actually making any phone calls.

How to run

Start the dialer

cd dialer
npm install
npm run dev

Start the visualization

cd vis
npm install
npm run dev

Usage

The dialer exposes it's data in two formats. The console, and an API to which the visualization can connect. With both services running, go to https://localhost:3000 in your browser to see the visualization.

How to contribute

1. Clone the repository and create a new branch
2. Create a new IDialer implementation (e.g. copy SerialDialer or PredictiveDialer) and give it a nice name. If you can't think of a better name, name it after yourself, e.g. ClaraDialer.
3. Experiment with the algorithm and look at the KPIs in the visualizer.
4. Once you're happy with the outcome, create a PR on this repository
5. In your PR, include an explanation of the algorithm as well as the benchmark KPIs from one testrun.

You will need to change dialer/index.ts to pick your dialer, and you might want to change the number of callers and targets as well while experimenting. But please do not commit those changes, and use the standard 10 callers, 1000 targets setting for the benchmark in the PR.

Objectives

Primary objectives

The primary objective is to achieve efficient calling, which translates to trying to find a good balance between the following (sometimes contradictory) KPIs.

• Dials per hour/caller as high as possible
• Drop rate as low as possible
• Average caller idle rate as low as possible
• Awkward silences (>1s) as low as possible

Secondary objectives

• Configurable "aggressiveness", e.g. by configuring accepted drop rate