SDK-from-csv

SDK: run multiple conversations from csv file

Description

A single Dasha application can run many conversations.
Moreover, it can concurrently run them, i.e. several conversations at a time.

There are many ways to store and pass conversation inputs into an application. It depends on the desired architecture and other constraints.

The current example demonstrates the approach for running multiple conversations with inputs described in a .csv file.

1. Here we've got a single Dasha application (with concurrency set to 3) and two .csv files which are passed to the app (asynchronously) as conversation inputs.
2. Conversation inputs are transformed via user-defined schema described in inputSchema.js (it is necessary since .csv file stores string data and the app may require some specific data type).
3. All results are written into single output .csv file. The conversation outputs are transformed into strings via user-defined schema outputSchema.js

Note: You can use this example as a template for your application.

Project description

• app - folder with Dasha files (dialogue script, nlu config, etc...)
• CsvRunner.js - a utility class used to handle csv file inputs
• index.js - main script
• inputSchema.js - schema for csv-input transformation (from string to desired type)
• outputSchema.js - schema for dialogue output transformation (from output type to string)
• input*.csv - .csv file with conversation inputs

Installation

1. Setup environment (see doc)
2. Run npm i in a current folder.

Running the demo

Run npm start.

Note that you may change the endpoint parameters in .csv files. For instance, you may set your phone number as an endpoint and Dasha will call that number. But, please, watch the _channel as well. If you use your phone number as an endpoint, _channel must be set to audio. Use chat as _channel value to start your conversation in chat.