# AI Platform Scorecard CLI

## PURPOSE
The video ends with a blunt claim: every platform on the market is "selling you an
engine with a seat bolted onto it and calling it a car," and buyers need "some kind of
scoreboard of your own." This Click + Pandas command-line tool is that scoreboard. It
ships with the transcript's own numbers — Anthropic Co-work 3/7, Google 4/7, OpenAI 5/7,
Microsoft 5.5/7, xAI Grokbot 1.5/7, Kylon 6.5/7, Victor AI 5.5/7, Lindy 5/7, Open Claw
2.5/7 — broken down across the 7 platform components, and lets you rank and filter them
from your terminal.

## WHY THIS MATTERS
It teaches that market data is only useful when it is queryable. Instead of reading a
review, an agency or buyer runs `scorecard rank --camp startup` and sees that the
startups (Kylon 6.5, Victor 5.5) are "ahead of the big labs," exactly as the video says,
then filters on the specific component they cannot live without.

## HOW TO BUILD
Click provides the command group; Pandas loads a bundled CSV and does the ranking and
pivoting. Key snippet for the per-component view:

```python
df = pd.read_csv(DATA_PATH)
comp = df.set_index("platform")[COMPONENTS].T
print(comp.loc[[component]].T.sort_values(component, ascending=False))
```

Build three commands: `rank` (sort by overall score, optional `--camp lab|startup`),
`components --name Kylon` (show the 7-piece breakdown), and `budget --max-per-seat`
(filter the per-seat price, e.g. Google at $7 or the free single-player Open Claw). Use
`click.echo` and a plain-text table so output pipes cleanly into other tools.

## WHEN TO USE IN PRODUCTION
Perfect for agency analysts doing vendor triage, for a founder who wants a repeatable
pre-sales checklist, or inside a CI job that flags when a new platform row is added.
Do NOT use it as a live pricing source — seat prices and scores change monthly (the
video notes the stack "keeps changing under you" and only relevant for that moment).
Not a scraper; treat the CSV as a versioned snapshot.

## HOW IT HELPS STUDENTS
You learn Click command/option design, argument parsing, Pandas indexing and pivoting,
reading packaged data files, and formatting CLI output. It is a realistic "internal
tooling" pattern every data team builds, and it reinforces the transcript's advice to
compare vendors component-by-component rather than on a single headline number.

## STARTER CODE
See `starter/main.py` — a runnable Click group with `rank`, `components`, and `budget`
subcommands and a deterministic embedded dataset.
