
nice work on this. one thing on the parser breaking if NAFDAC changes their layout though. you could let the parser run first since it’s cheap and fast, and if it fails or the data comes back weird (missing fields, bad formats etc), do a quick check on the raw response. if it’s just an error page or rate limit you log it and move on. but if the HTML looks like real content and the parser still choked, that probably means the structure changed so you fall back to an LLM to extract the fields as JSON. validate the output either way cos LLMs will make stuff up. and every time the LLM fallback fires, log it and flag your team so someone updates the parser. LLM just buys you time, parser stays the primary path. means your API stays up even when NAFDAC eventually decides to redesign.
English


















