Oddly, your question is at the intersection of search and DevOps, areas of functionality that Airtable has never demonstrated competence.
I recently built a system that tracks 600 bases. Confidentiality prohibits me from exposing the details, but read this, and then consider these points.
- Each base is about something. Its meta information is extractable through several attributes, including but not limited to table names, descriptions, etc.
- AI embeddings help you instantiate meaning in a natural language context using whatever attributes and terms you care to collect.
- With a vector space of all bases, you now have a highly-focused knowledge base of all bases.
Vector spaces can be used for search using natural language. They can also be used for filtering like any other data store. And they can even be used in a hybrid fashion - e.g. …
“Show me all bases about accounting transactions; filter on keyword A/R”
Who here has religiously documented what bases and tables are about using the native Airtable fields? LOL. Now you have reason to do exactly that.