Articles · Gemini AI Tools
Cache a large source once, then ask again without pretending free-tier caching works
A long-document page is easy to treat as a free unlimited memory for every PDF you will ever paste. It is not. Long-Context / Cache Demo creates an explicit Gemini context cache (cachedContents) when this host is paid and configured, then asks follow-up questions against that cache. Repeated questions about the same large content may use Gemini context caching to reduce processing cost. The free tier fails closed. Short sources are refused before a pretend cache is created. Temporary cache ids are tracked in memory and deleted when you clear the page or when cleanup expires them.
Why this is worth doing in the browser
People want context caching because re-sending a huge transcript for every question is wasteful. Shipping a free-tier “cache” that is really a full re-prompt would be dishonest. Related Long Document Analyzer still helps for one-shot analysis without an explicit cache. Related Batch Processor is for many independent prompts, not one shared source. Answers are drafts. Check names and numbers against the source.
How to use the tool
Paste a large source. Confirm cloud caching. Create cache on a paid host. Ask questions. Delete cache or Clear / Start Over removes the temporary entry.
- Do not paste confidential sources unless you accept Google processing.
- Explicit caching requires a paid Gemini configuration on DevNestro.
- A short paste is refused instead of faking a cache hit.
- Delete the cache when you leave a shared computer.
Privacy
The source and later questions are sent to Google Gemini. Cache names are temporary and are deleted when you clear the page or when cleanup runs. This is not processed only in your browser. Free-tier training language follows the configured tier.
Have a large source you are willing to send on a paid host? Open Long-Context / Cache Demo, create the cache once, and ask focused questions against it.