Skip to content

ttzeng/chat-bot

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

12 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

My LINE chat bot

Overview

This repository contains the source code of my LINE chat bot, leveraging Dialogflow, LINE Messaging API, Open Data API from Central Weather Administration (中央氣象署), and OpenAI API technologies.

The chat bot uses Dialogflow to handle user chat messages. Dialogflow is a natural language understanding platform that provides a conversational user interface for integrating to mobile apps, devices, bots, and web services, etc. Diaglogflow translates end-user text or audio during a conversation to structured data that computer programs can understand through virtual agents, where agents match the end-user expression to the best combined intents (i.e. intent classification) to handle a complete conversation. Dialogflow also integrates with many conversation platforms such as Google Chat, Facebook Messenger, and LINE to handle end-user interactions in a platform-specific way. In addition to static responses, fulfillments can be enabled to call an user predefined service (i.e. webhook) to provide a more dynamic response.

How to Set up

  1. Before deploying this LINE chat bot, follow this guide to create a channel using LINE Developers Console - the portal to manage Developer, Provider, and Channel on the LINE Platform.

  2. Deploy the source code of this repository as a web service. Since the chat bot requires the access token (LINE_CHANNEL_ACCESS_TOKEN) and channel secret (LINE_CHANNEL_SECRET) created in the previous step, the authorization key (CWB_AUTHORIZATION_KEY) for accessing the Open Weather Data platform by CWA (中央氣象署), as well as the OpenAI secret API key (OPENAI_API_KEY) when invoking individual APIs, these secrets must be either provided in a local file named ".env", or manually configured as environment runtime variables for the web service. For example, the chat bot service is hosted on Google Cloud Functions as below.

  3. Follow this guide to create an agent using Dialogflow ES Console. The new agent initially has two default intents: Default Fallback Intent and Default Welcome Intent. The default welcome intent is supposed to be matched while starting a conversation with greetings, but it is removed afterward for this chat bot, since we are going to proxy the fallback query messages to OpenAI for Generative Pre-trained Transformers (a.k.a., GPT) responses.

  4. Click Entities in the left sidebar menu, and add "@getLocation" and "@getWeather" custom entities to match location (e.g. "台北") and weather (e.g. "天氣", "晴天") specific keywords. These entities will be used during the training annotation in the next step, to extract the end-user expression to intent parameters.

  5. Back to the Intents panel, create a "Query Weather Intent" and train the intent with phrases that users typically use to query whether data. Annotate the given phrases with system and/or custom entities created in the previous step, to train the agent for extracting parameter values.

  6. Click Fulfillment in the left sidebar menu, enable Webhook feature and enter the trigger URL of the web service created in Step 2 to the URL field in the panel, then turn on "Enable webhook call for this intent" in the Fulfillment section of both "Query Weather Intent" and "Default Fallback Intent", the web service will be called by Dialogflow while matching the named intents.

  7. Click Integrations in the left sidebar menu, select the LINE integration box, mark down the Webhook URL on the LINE integration pop-up window, and enter the URL to the Webhook settings of the LINE channel created in Step 1 to complete the setup.

Resources

About

Chat bot 聊天機器人

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages