Perhaps you could save some money by telling the AI to write a Python code that would scrape some information from the websites, convert them to plain text (much less data), and then using the AI to process that text?
The idea is that you pay for creating the Python code, and for processing its output, but you don’t pay for the Python code downloading and processing the data.
I’m not all that sure how AI search works. Searches, and indexes top 20 hits, or something like that. Is reading a webpage the expensive part? If so then caching/context window management might matter a lot. Plain text might backfire if you actually lose table structure and stuff. You can probably ignore styles at least.
Perhaps you could save some money by telling the AI to write a Python code that would scrape some information from the websites, convert them to plain text (much less data), and then using the AI to process that text?
The idea is that you pay for creating the Python code, and for processing its output, but you don’t pay for the Python code downloading and processing the data.
I’m not all that sure how AI search works. Searches, and indexes top 20 hits, or something like that. Is reading a webpage the expensive part? If so then caching/context window management might matter a lot. Plain text might backfire if you actually lose table structure and stuff. You can probably ignore styles at least.