ReturnVisitRobot

Insurance callback robot

This is an outbound call robot used for insurance return visits. Only the core text robot part is retained in this project. The core functional process is as follows:

This project rewrites the dialogue response strategy based on finite state machines based on rasa. It should have some reference significance for friends who want to build text robots.

User Guide

Configuration, you can modify the pipeline in the confg.yml file. For the specific configuration method of the pipeline, please refer to the Rasa official website https://rasa.com/docs/rasa/model-configuration

The following is a configuration example of two custom components:

Memory-Based Dialogue Response Strategies

 policies:
  - name: policy.memoization.MemoizationPolicy
  - name: TEDPolicy
    max_history: 20
    epochs: 15
    batch_size: 50
  - name: RulePolicy
    core_fallback_threshold: 0.3
    enable_fallback_prediction: True
    core_fallback_action_name: "action_default_fallback"

Or based on finite state machine (FSM)

 policies:
  - name: policy.fsm_policy.FsmPolicy
  - name: TEDPolicy
    max_history: 20
    epochs: 15
    batch_size: 50
  - name: RulePolicy
    core_fallback_threshold: 0.3
    enable_fallback_prediction: True
    core_fallback_action_name: "action_default_fallback"

train

 python main.py train

run

 #开启动作响应服务器，默认5055端口
python main.py run actions
#开启对话shell,如果使用pycharm注意勾选emulate terminal
python main.py shell
#第一句请输入内置意图：开始
Your input:开始

Prerequisites project environment dependencies

 six~=1.16.0
rasa~=3.1.0
zlib~=1.2.12
tqdm~=4.64.0

Install

 pip install -r requirements.txt

Deployment method, which can call the session response through rest mode

 rasa run -m xx.tar.gz(模型文件) -p 8787(端口号)