New: Cookbooks and AI ExplanationsStep-by-Step recipes to solve problems connected to Roadmaps and Cheat Sheets. Need more details? Use AI buttons for structured and simple explanations with concrete examples throughout the whole platform.Take a look
Your vector database has one bad five minutes, and your whole app goes down with it.
What you'll have at the end
A tested fallback path that keeps answering when one dependency times out, with a log line every time it fires
You need
A RAG app already retrieving from a vector database in production, plus the exact same chunk text available somewhere else you can also read from, such as the file or table you originally loaded into the vector database.
Not covered
Blending keyword and vector results together for better everyday relevance, or building a second provider for the model that generates the answer: this only switches retrieval to keyword search while the vector database itself is unreachable.
Leans on
RAG's vector search misses exact terms that hybrid search catches
Go there when the problem is everyday relevance rather than an outage: that recipe blends both searches on every request, while this one only switches to keyword search while the vector path is actually unreachable.
Your app breaks the moment your LLM provider does
Use this instead when what's failing is the model generating the answer rather than the database retrieving what it answers from; the failure signal and the fallback both sit one layer over from this recipe's.
Put every model call behind one gateway that routes, caches, and falls back
Pair with this once both exist: it gives every model call the same fast-fail-and-switch shape this recipe just gave retrieval.
Checked 18 Aug 2026
Part of the AI Engineering cookbook