API Maker
The backend engine for the AI era
DatabasesEpisode 2 5:46 11 chapters October 2026
Connect MongoDB, MySQL, MariaDB, PostgreSQL, SQL Server and Oracle to one API Maker backend. Connection strings stay in a secret on the server, every instance is tested before it is saved, and every table answers REST APIs without a line of code.
In the playlists The complete course Databases & schemas
MongoDB, MySQL, MariaDB, PostgreSQL, SQL Server and Oracle, connected to one backend, with every table answering APIs. In this video, the warehouse of Calm Home joins in, live.
Six databases usually mean six drivers, six pools, six URL formats, passwords in .env files, and too often in Git, and CRUD code written table by table. With API Maker, one form per database server, its connection string kept in the default secret, on the server, and the APIs of every table, without a line of code. In this video : the default secret, an instance added and tested, databases and tables, schemas for every table at once, the APIs that appear, and the life of an instance.
Security, Secret Management. The secret Default holds what must never reach Git. Under connectionString, one key per engine : mongodb, mysql, mariadb, sqlServer, postgresql and oracle, with its username and password. The values stay on the server, encrypted : here they are blurred. Nothing to change. An instance never holds a connection string. It names a key of the secret, and that name is what goes to Git. Each server reads the value from its own secret : the same instance works on every environment.
APIs and Logic, Instance API. Add New. Six tiles for eight databases : TiDB and Percona XtraDB connect as MySQL. The SQL Server tile : the key sqlServer of the secret is picked by itself, and the form shows how its value looks. The name, sqlserver, is part of every API address. Test Instance : connected. Save Instance : it tests again, then saves. The new instance lists the databases of the server, with their size.
warehouse, and its three tables : inventory, shipments and warehouses.
Bulk Schema Creation lists the tables without a schema. All three, Generate Schemas : each schema is written from what the table holds today.
dbo.inventory : one hundred and forty four rows, one index, and its APIs are there. Twenty eight : fourteen SCHEMA APIs, checked against the schema, and fourteen GEN APIs, straight to the table. A MongoDB collection gets seventeen of each. Nothing was written. Every engine looks the same here. MongoDB, catalog and its collections. PostgreSQL, crm : its tables carry their schema name, public. Oracle, HR, with upper case names.
One page, one way, six engines.
And a database that is not there yet is made from this page too. Add New Database, its name, and the statement written for its engine, shown before it runs. Saved. Its tables come from a designer, with their schemas : video fifty four shows it. One more tool, for MongoDB. The database catalog, Compact Database : every collection with its documents, its storage, and the space MongoDB still keeps from removed documents, reusable.
Here, old review drafts were cleaned up. Compact asks first. Yes : each collection is rewritten on its own, a failure never stops the others, and Space Freed says what the files gave back. No shell, and no command typed for each collection. A whole instance at once, too. The instance mongodb, More Actions, Compact Instance : every database of the instance, one after the other. Here, catalog, and shop_events, where old page views were removed.
admin, config and local, the databases of MongoDB itself, are left alone. Compact, Yes : what each database gave back, and the total.
More Actions, Clone : sqlserver-clone. On the clone, Make Inactive : its APIs stop until it is active again. Show All brings it back, greyed. Permanently Delete : its schemas, hooks and settings in API Maker go with it ; the data in the database is not touched.
The APIs of an inactive instance answer four hundred, as if it did not exist. A rename follows every reference : hooks, groups, settings and logs. Oracle adds two keys, the username and the password, and a privilege. And the key of any text of the secret works : connectionString is only the convention.
No credential in code or in Git : only the name of a key. The same API shape and the same errors on every engine. So an AI assistant asked to connect a database can not leak a password into the repository.
To sum up. Connection strings in the default secret, on the server. An instance per database server, tested before it is saved. Schemas for every table, in one click. Every table gets its APIs, on six engines. Next : schemas in depth. API Maker. Build faster, spend smarter, deliver better.
Length 5:43Schemas
Length 11:06Table designer
Length 5:08Data types
Length 5:34Data safety