rip-cd

Files (17)

• SKILL.md
• eval_queries.json · evals/
• evals.json · evals/
• check-deps.sh · scripts/
• eject-disc.sh · scripts/
• encode-release.sh · scripts/
• fetch-cover.sh · scripts/
• accuraterip.py · scripts/lib/
• disc.py · scripts/lib/
• out.sh · scripts/lib/
• lookup-release.sh · scripts/
• make-release-record.sh · scripts/
• read-disc.sh · scripts/
• rip-disc.sh · scripts/
• update-record.sh · scripts/
• verify-accuraterip.sh · scripts/
• work-dir.sh · scripts/

SKILL.md (first lines)

## Why not Apple Music's own import

Apple Music's CD import names tracks from its own lookup, which for many
discs repeats the release name on every title ("Intro (live at Foro Sol
(Mexico City) 21/23/25.09.23)"), files each disc of a set as its own album
("M" and "M [Disc 2]"), guesses artwork that is often for a different album,
and never checks the rip or keeps a lossless copy. This skill does all of it
itself. **If Apple Music starts importing the disc on its own, tell the user
to cancel that import first**; to stop it for good, Music > Settings >
General > "When a CD is inserted" > "Show CD".

## Setup

```bash
bash ../setup/scripts/check-config.sh
```

Needs `library_import`, `library_storage` and `archive_workdir`. Uses
`cd_read_offset` (the drive's read offset in samples) and `work_dir` (where
temporary work goes) when set. If a
required key is `missing`, stop and run the `setup` skill.

Then check the tools:

```bash
bash scripts/check-deps.sh
```

If `ok` is false, **stop**. Tell the user which tools are missing and the
`install_command`, and offer to run it. Run it only when they say yes; never
rip with a tool missing.

## Running the scripts

Scripts are in `scripts/` in this skill's base directory. Run each one as a
single command by its **absolute path**, with no redirect, pipe or `&&`, and
save JSON with the script's own `--out FILE`:

```bash
bash /abs/path/to/skills/rip-cd/scripts/read-disc.sh --out "$WORK/disc.json"
```

The plugin's safety hook auto-approves exactly that shape; anything else
(`> file.json`, `| jq`, a relative `scripts/...` path) asks the user every
time. Below, `scripts/x.sh` is shorthand for that absolute path.

Work in one temporary folder per release, `$WORK`, from
`scripts/work-dir.sh "<name>"` (its `work_dir`): `cd-rip` while the release
is unknown, `cd-rip-<Artist> - <Album>` once a set's first disc is chosen
(keep using whichever folder the set started in). It lives under the
system's temp folder, never in Downloads or the user's home: nothing in it
is the only copy of anything once a disc is archived, because the release
record and cover go to the NAS with the WAVs.

## Scripts

- **`work-dir.sh <name>`**: the temporary work folder for a release.
- **`check-deps.sh`**: are cd-paranoia, ffmpeg (with LAME), python3, curl and jq installed.
- **`read-disc.sh`**: the disc's table of contents, its MusicBrainz disc ID and AccurateRip IDs (computed locally), and any names macOS shows for it, with repeated suffixes found.
- **`lookup-release.sh <disc.json>`**: every MusicBrainz release that disc belongs to.
- **`make-release-record.sh --release-id ID --genre G [...]`**: the release record (`release.json`) all later steps read.
- **`fetch-cover.sh <release_id> <cover.jpg>`**: the front cover from the Cover Art Archive, padded square.
- **`rip-disc.sh <dir> --offset N`**: secure rip to `01.wav`, `02.wav`, ...
- **`verify-accuraterip.sh <dir> <disc.json>`**: checksums against AccurateRip.
- **`encode-release.sh <dir> <release.json> <disc> <out> --cover cover.jpg`**: `mp3/` (320 kbps, tagged, cover embedded) and `wav/` (the rip, untouched, renamed, plus release.json and cover.jpg).
- **`update-record.sh <release.json> --disc N [--accuraterip S] [--discid ID] [--title N=T]`**: record a disc's verdict, ID or a corrected title, in the record and its copies in the work folder.
- **`eject-disc.sh`**: eject the disc.

Every script takes `--help`.

## Workflow

### Step 1: Identify the disc

```bash
bash scripts/read-disc.sh --out "$WORK/disc.json"
bash scripts/lookup-release.sh "$WORK/disc.json" --out "$WORK/lookup.json"
```

Exit 2 from read-disc means no disc: ask the user to insert one.

As soon as the lookup returns, **tell the user what the disc is**: artist,
album, which disc of how many, and track count ("This is Night Drive -
Afterglow Live, disc 1, 3 tracks"). When several releases carry the disc,
say so in the same breath and that the barcode on the back of the case tells
them apart (Step 3, question 1). Every message that asks a question repeats
the disc's name, so the user never has to scroll back to know which disc
the question is about.

**Is this a later disc of a release already started?** Look for a
`release.json` whose discs list this disc ID in `known_discids` or
`mb_discid`: in `$ARCHIVE_WORKDIR/to_nas/wav/*/*/release.json` (Glob) and in
the work folders beside `$WORK`. If one matches, say so ("this is disc 2 of
M, which you ripped on 19 Oct"), **reuse that record's answers** (no
questions in Step 3) and record the disc's ID:
`scripts/update-record.sh "$RECORD" --disc N --discid "$DISCID"`. When the
match is on the NAS only, copy that record into `$WORK` with the Read and
Write tools first. Every disc of a release must carry
identical Album, Album Artist and disc total, or Apple Music splits the
release into several albums.

### Step 2: Start the rip

The rip needs no answers, so start it now and ask the questions while it
runs (15-25 minutes for a full CD). Run it in the background
(`run_in_background`), and tell the user it is running:

```bash
bash scripts/rip-disc.sh "$WORK/rip-disc$N" --offset "$CD_READ_OFFSET" --out "$WORK/rip-disc$N.json"
```

`$N` is which disc of the release this is; until the release is chosen, use
the candidate's `audio_disc`. "405: Option not supported by drive" in its
output is the drive declining an optional speed command and is harmless.

If `cd_read_offset` is not configured, say the AccurateRip check will
probably fail until it is, and rip with 0: the rip is still secure. The
offset for a drive model is in AccurateRip's drive list; once a rip verifies
with it, add it to config.json as `cd_read_offset`.

From eyelock/assistants, MIT.

More from eyelock/assistants