- Added a required Pronouns dropdown to /request and /admin/request:
He/Him/His, She/Her/Hers, They/Them/Their.
- Moved Email to the top of the customer request form; Pronouns sit just below Name.
- Added column to the requests schema with automatic migration.
- Stored pronouns are included in confirmation emails and Hermes prompt copy.
- Fixed /admin/request so Preferred singer voice / gender is a dropdown
matching the customer form instead of a free-text input.
- Updated README to describe pronouns and the structured style selectors.
Bumps version to 0.4.7.
The customer request form previously had an open 'Style / genre / mood'
text field, which produced inconsistent genre descriptions and confused
Suno. Replace it with three dropdowns:
- Decade / era (required) — loaded from /mnt/Storage/Decades.txt
- Basic style (required) — loaded from /mnt/Storage/Music Genres.txt
- Additional style (optional) — same genre list
The combined value is stored in the existing style_genre column as a
comma-separated string (e.g. '1980's, Pop, Funk'), so no schema change
is required. The admin request page exposes the same dropdowns for
corrections, and the copy-to-Hermes prompt formats the style as a clean
sentence like '1980's-era Pop with Funk influences'.
The canonical lists live in /mnt/Storage; bundled copies under lists/
act as a fallback when that mount is unavailable in the container.
Bumps version to 0.4.6 and updates README.
- Clarified HERMES_API_KEY and PRICE_PER_VERSION comments in config.py.
- Completely rewrote README.md with:
- version, overview, and table of contents
- page-by-page customer and admin guide
- status flow table
- detailed /admin/settings section
- Docker / Portainer install process
- full environment variable table
- local development and troubleshooting sections
The signed callback token for /api/prompt/<rid> expired after 1 hour.
That was too short for a workflow where the operator may copy the URL and
paste it to Hermes, then wait for a response. Extend to 7 days.
Also fixes the misleading 401 on expired tokens (token check runs before
API key check). The new /api/key-test endpoint lets callers verify the
API key independently.
Bump version 0.4.4 -> 0.4.5.
Provides a simple GET endpoint that validates the Authorization Bearer
token against HERMES_API_KEY and returns JSON:
{"ok": true, "reason": "valid"} or
{"ok": false, "reason": "key_mismatch" / "missing_bearer" / "not_configured"}
Rate limited to 4 per minute to prevent brute-force guessing.
Bump version 0.4.3 -> 0.4.4.
The previous schema migration added revision_count as TEXT. SQLite stores
the default as TEXT, but the app does integer arithmetic on it, causing:
TypeError: unsupported operand type(s) for -: 'int' and 'str'
- init_db() now adds revision_count as INTEGER.
- app.py casts revision_count to int in play() and revise() and when
incrementing in revise(), so existing TEXT values also work.
Bump version 0.4.2 -> 0.4.3.
Deployed DBs persisted without the revision_count column / revision_history
table, causing customer revision requests to 500.
- init_db() now migrates existing tables by adding missing columns.
- log_revision() falls back to init_db() on missing table for full repair.
- Bump version 0.4.1 -> 0.4.2.
Deployed databases can persist without the revision_history table,
causing /play/<token>/revise to crash with sqlite3.OperationalError.
- log_revision() now creates the table on the fly if it is missing.
- app.py runs init_db() at import time (safe CREATE TABLE IF NOT EXISTS)
so new deployments auto-create missing tables on startup.
Bump patch version 0.4.0 -> 0.4.1.
/api/prompt/<rid> previously rejected any request not in 'pending'
status. This blocked the revision workflow from using the callback.
Now the endpoint accepts 'pending' (sets status to 'prompt_ready') and
'revisions_requested' (keeps status as 'revisions_requested' so the
operator still sees it needs new song uploads).
When a request has revision_note set, the operator now sees a
'Copy revision prompt for Hermes' button in the revision note box.
It copies a prompt that includes:
- the original customer brief,
- the previously generated title/style/lyrics,
- the customer's revision request,
- the signed callback URL and expected JSON shape.
Minor version bump: 0.3.0 → 0.4.0.
- New VERSION file at project root, starting at 0.3.0.
- config.py reads VERSION and exposes it as Config.VERSION.
- Admin settings page now renders the live version from config.
- Semantic versioning convention: MAJOR for breaking/user-facing changes,
MINOR for new features, PATCH for bug fixes and small polish.
Some rows (schema drift, cancelled requests, manual status edits) have
customer_approved = NULL instead of the default 'none'. Templates called
.upper() on it blindly, causing a Jinja2 UndefinedError / 500 when the
customer player or admin pages loaded.
Replace all .upper() calls with (value or 'none').upper() and add truthy
checks before rendering the approved choice in status.html and
dashboard.html.
The customer player page (/play/<token>) and revision endpoint were
reading max_revisions directly from booth_settings.json, where it is
stored as a string. Arithmetic/comparison with revision_count (an int)
raised TypeError and caused a 500 error when opening the customer player.
Add get_max_revisions() helper that always returns a non-negative int,
and use it in play(), revise(), and admin_settings().
The regenerate button stores encrypted keys, but legacy/manual writes may
save plaintext. get_hermes_api_key now falls back to returning the raw
value if Fernet decryption fails.
Previously passed the entire runtime_settings dict as metadata, which could
clobber/overlap email form fields. Now only artist/album/year/comment are passed.
- Disable checkboxes, payment ref input, and submit button in admin UI when customer has not approved.
- Add server-side guard in mark_paid_deliver action.
- Replace /audio/<token>/<v>.mp3 with /api/stream/<token>/<v>.mp3.
- Stream bytes from private storage with Accept-Ranges and Content-Range support.
- Keep old /audio route returning 404.
- Update player page JS to use new stream endpoint.
