add-code-units

Add Local Source Code

On Entry

Tell the user:

Importing local SQL files.

I'll handle this for you — importing your local SQL files and organizing them by type into source/, ready for conversion. It's a single automated step; you won't need to copy, arrange, or move anything yourself.

Prerequisites

• Migration project initialized (scai init)
• Local directory containing SQL source files

Workflow

Step 1: Get Source Path

Ask the user for the path to their source SQL files:

"Where are your SQL source files located? Please provide the full path to the directory."

Step 2: Annotate Multi-Object Files with sc-tags

scai code add uses the arrange engine to split files into one object per code unit. For non-T-SQL dialects the engine splits on /* <sc-...> */ boundary tags. Before importing, call the split_code MCP tool to detect files that contain more than one top-level object and inject those tags automatically.

split_code(input_path="<INPUT_PATH>")

• If all files are single-object, the tool says so and you can skip ahead to Step 3 using <INPUT_PATH> directly.
• If multi-object files are found, the tool copies all .sql files to artifacts/source_split/, injects the sc-tags, verifies the result, and tells you to use artifacts/source_split as the input for Step 3.

The tool handles the full detect → inject → verify cycle in one call. It is OS-agnostic and does not require Python.

Example — what the tool injects before each object in a multi-object file:

/* <sc-table>sales.public.customers</sc-table> */
CREATE TABLE sales.public.customers (
  id INT,
  name VARCHAR(100)
);

/* <sc-view>sales.public.v_active_customers</sc-view> */
CREATE VIEW sales.public.v_active_customers AS
SELECT * FROM sales.public.customers WHERE active = 1;

T-SQL note: T-SQL files do not need this step — the engine has a native T-SQL splitter. Skip Step 2 for T-SQL source files.

Step 3: Add Code to Project

Use the path returned by split_code (or <INPUT_PATH> directly if no multi-object files were found):