The no-API constraint on the ticket system is what tripped me up at first.
Start by clarifying requirements and constraints, then propose a hybrid architecture that separates the browser-only ticket data from the API-accessible user data. Focus on how to integrate these sources securely and efficiently, and discuss trade-offs like caching, data consistency, and real-time updates.
Pro tip: Emphasize the importance of decoupling the ticket system from the user API to avoid tight coupling and enable independent scaling. Also, mention the need for a secure proxy or backend-for-frontend to handle browser-only data and API calls, ensuring compliance and performance.
Ask questions to understand the scale, real-time needs, security requirements, and why ticket data is browser-only. Confirm that user data API is the only way to fetch user info and generate chat responses.
Propose a system with a frontend for agents, a backend service (BFF) that interacts with the browser-only ticket data (e.g., via scraping or browser automation) and the user API. Consider using a message queue for asynchronous processing.
Design how the backend will combine ticket data (from browser) and user data (from API) to generate chat responses. Discuss using a template engine or AI model, and how to handle latency and failures.
Address scaling the browser automation (e.g., headless browsers), caching user data, rate limiting, and securing access to both data sources. Mention monitoring and logging.
Discuss trade-offs of browser automation vs. asking for API access, caching vs. real-time, and monolithic vs. microservices. Suggest potential improvements if constraints change.
AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.