CEO of AI company “Predictive Technology”, Mr. Pan, shares the practice of intelligent question generator. Common intelligent agents are mostly in search-like scenarios, such as knowledge base query and customer service. A friend wants to create an intelligent question generator agent that requires maintaining a question bank. When users take an exam, the agent randomly generates questions and evaluates and scores them. Connecting a database or external interface can achieve scoring management. There are some pitfalls when using the no-code tool Coze: 1. The knowledge base requires manual data cleaning and cannot directly import large documents. 2. Simple prompts cannot achieve active random extraction of content from the knowledge base as in non-workflow mode, search needs to combine user instructions and context. 3. Switching to workflow mode. The design idea is that when a user initiates an exam request, it triggers the workflow. The workflow randomly searches for questions in the knowledge base to ensure that the questions in the exam do not repeat. Each question is given a unique number. Add a code module to randomly generate an id and ensure it is not in the array of already generated question ids. When taking the exam again, the array is emptied. 4. When writing prompts, draw on the safety protection experience of the red envelope intelligent agent and design a safety code. You can call an interface to write scores to the management system or integrate with the access control system. In summary, to achieve the function of randomly extracting knowledge fragments, you need to call a workflow to randomly request question numbers from the knowledge base. You can also build a program that randomly finds questions and let the workflow call its API to return questions. The AI can adapt the question types or options. If you have a better method, please leave a message. You can follow, like, forward and join the community to understand the dynamic knowledge of AI.
Posted inAI News
