aahirsch / df-custom-tools

A repo of custom tools for use with DialogFlow CX

Scripts for the Wharton Behavioral Lab's Negotiation Bot

map_routes.py

Converts a Dialogflow agent's dialog taxonomy to an easily-viewable Excel spreadsheet. Works by using Google's Dialogflow CX Python client to scrape an agent's pages in a particular flow.

To use:

• Download a credentials file from the Google Cloud Platform console.
• Run export GOOGLE_APPLICATION_CREDENTIALS="<insert credentials here>" in your terminal
• Set the AGENT environment variable to the agent whose routes you'd like to map. The content should be in the format projects/.../locations/.../agents/...
• Set the FLOW environment variable to the specific flow whose pages you'd like to scrape the routes from. The content should be in the format {}/flows/...

create_variant.py

Enables mass editing of a Dialogflow agent's dialog via an Excel spreadsheet. Works by using a route map (see map_routes.py) to parse changes to an agent's fulfillments; then uploads the generated "variant" as a new Dialogflow agent.

To use:

• Run map_routes.py on an existing agent.
• Edit route fulfillments as you'd like. Keep in mind that only fulfillments in the edited route map will be uploaded to the generated variant agent.
• Set the ROUTE_MAP environment variable to the file path of the edited route map.
• Set the VARIANT_NAME environment variable to the name of the new bot.

prioritize_intent.py

Prioritizes a given intent by setting its internal priority value to an inputted integer. See intent priorities.

To use:

• Set the INTENT environment variable to the intent whose priority you'd like to change. The intent should be in the format projects/.../locations/.../agents.../intents/...
• Set the NEW_PRIORITY environment variable to the integer value you'd like.

sync_training.py

Syncs training phrases for matching intents between two agents. Works by examining all common intents between a reference agent and a target agent, and then updating the target agent's training phrases with the reference agent's.

To use:

• Set the REFERENCE environment variable to the agent whose training phrases you'd like to copy over. The content should be in the format projects/.../locations/.../agents/...
• Set the AGENT environment variable to the agent whose training phrases you'd like to update. The content should be in the format projects/.../locations/.../agents/...

paraphrase_text.py

Paraphrases input text using Google Research's PEGASUS model.

To use:

• Set sentence variable to an input sentence you'd like.