基于Rasa构建聊天机器人：Ubuntu本地环境下对话数据获取问题

Hey there! I’ve dealt with this exact situation before when testing Rasa bots locally, so let me share a few straightforward ways to grab your conversation data—no need to deploy to a full HTTP server setup.

方法1：直接读取Rasa默认SQLite数据库 #

When you run rasa shell locally, Rasa automatically stores all conversation events (user messages, bot responses, intent predictions, etc.) in a SQLite database file right in your working directory. Here’s how to access it:

• First, confirm the database exists: Run ls -l in your bot’s root folder—you should see a file named rasa.db.
• Install SQLite CLI (if you don’t have it):

  sudo apt update && sudo apt install sqlite3
  

• Access the database and query events:

  sqlite3 rasa.db
  

  Once in the SQLite shell, run this query to see all conversation data:

  SELECT * FROM events;
  

• Export to a readable format (like CSV):
  Exit the SQLite shell, then run this command to save events to a CSV file:

  sqlite3 -header -csv rasa.db "SELECT * FROM events;" > conversation_data.csv
  

方法2：配置文件型事件日志 #

You can set up Rasa to write all conversation events directly to a JSON log file. Just add this to your config.yml:

event_broker:
  type: file
  path: ./conversation_events.log
  format: json

Restart your bot with rasa shell, and every time a conversation happens, the events will be appended to conversation_events.log in JSON format. You can parse this file with tools like jq (install with sudo apt install jq) to filter or format the data:

jq '.text' conversation_events.log  # Extract only user/bot messages

方法3：启用本地API服务（匹配文档场景） #

If you want to use the exact workflow mentioned in the Rasa docs, you can run the API server alongside the shell:

1. Open a new terminal window, navigate to your bot’s folder, and start the API:

   rasa run --enable-api
   

2. In your original terminal, run the shell connected to the local API:

   rasa shell --endpoint http://localhost:5005
   

Now you can use the standard conversation tracker endpoints to fetch data—just send requests to http://localhost:5005 using tools like curl or Postman.

内容的提问来源于stack exchange，提问作者Vincent