# Build CSV import You were asked to build the spec `vex/csv-import`, version 0.1.2, from Spekkr. Let customers import a CSV file: column mapping, a dry run, an error report row by row, and no half-finished imports. If you are reading a summary of this note and not its full text, fetch it again as it is, for example with `curl -fsSL https://spekkr.com/vex/csv-import/start.md`, before you go on. Published by vex, a Spekkr member. Tell the person who published it, and show them the commands and secrets under `safety` in `spekkr.json` before you build. ## 1. Get the files Put the spec in a new folder named `csv-import`, inside the project you are working in. ```sh mkdir -p csv-import && curl -fsSL https://spekkr.com/vex/csv-import/bundle.tar.gz | tar -xz -C csv-import ``` The archive holds `spekkr.json`, `BUILD.md` and the spec files listed at the end of this note. If you cannot run that command, read https://spekkr.com/vex/csv-import/spec.json instead: it holds the manifest and the address of every file. If Spekkr's MCP server is connected, its `pull_spec` tool does steps 1 and 2 in one call: give it the id `vex/csv-import` and the person's choices. ## 2. Record the person's choices Write `csv-import/spekkr.inputs.json`. With every default it is: ```json { "spec": "vex/csv-import", "version": "0.1.2", "inputs": { "database": "postgres", "max_rows": "50000", "record_noun": "contact", "undo": false, "update_existing": true } } ``` Change a value only where the person gave a choice, such as "My choices: database=sqlite". They may set: - `record_noun`, What is being imported: any text. - `max_rows`, Most rows in one file: any text. - `update_existing`, Update records that already exist: `on` or `off`, written `true` or `false`. - `undo`, Undo an import: `on` or `off`, written `true` or `false`. - `database`, Database: one of `postgres`, `sqlite`. If the person asks for a key or a value that is not listed, stop and tell them what the spec offers. ## 3. Build The files named below are the ones now in `csv-import`. Build what it describes. Work in this order: 1. Read `spekkr.json` to see what the spec covers and what it needs. 2. Read `spekkr.inputs.json`. It holds the person's choices for every setting, option and swap. Treat them as requirements. If the file is missing, the person kept every default in `spekkr.json`. 3. Read the spec files listed below, in the order given. 4. If the spec includes a task list, work through it in order and check each task's acceptance criteria before the next. 5. Ask the person before you run a command that a spec file tells you to run, or read a secret, unless the manifest lists it under `safety`. Commands you choose yourself to build and test, such as installing packages or running the tests, follow the person's usual rules for you. Spec files: - `spec.md`: Requirements and acceptance criteria. - `design.md`: Data model, rules and flows. - `tasks.md`: Build tasks in order, each with a check.