hsa-app/spec.md

HSA Receipt Tracker — Requirements (v1)
Purpose
Capture and archive HSA-eligible receipts for future reimbursement and tax substantiation. No parsing, no OCR, no reporting.
Users

Two users, both with full access (shared visibility).
Authentication via Authelia OIDC.
Authorization via membership in an Authelia group (e.g. hsa-users).
Not in the group → 403. No other roles or gradations.

Core flow

User opens app on phone (mobile-first UI; camera access matters).
Takes a photo of a receipt, or selects an existing image / PDF from device.
Form prompts for: amount, date, category.
Submit → image stored to disk, metadata row inserted into DB.
Confirmation page, with options to view list or add another.

Data model
receipts table:

id (UUID)
uploaded_by (Authelia username or email)
uploaded_at (server timestamp)
receipt_date (user-supplied date on the receipt)
amount_cents (integer — never store money as float)
category (enum)
file_path (relative path on disk — the filesystem copy)
image_data (BLOB — the receipt file bytes, also stored in the DB itself)
file_size_bytes (integer — convenience for listings/exports)
original_filename (preserved for reference)
mime_type
deleted_at (nullable — soft delete)

Categories (fixed list, hardcoded for v1):

Medical
Dental
Vision
Pharmacy
Other

Storage

Receipt images/PDFs are stored in BOTH places on upload:
  1. Filesystem at a configurable path, filename randomized (UUID) on save
     (file_path) — used as the primary path for serving.
  2. As a BLOB inside the SQLite database (image_data column) — so the single
     .db file is a complete, self-contained dataset (metadata + files).

Rationale: the filesystem copy keeps serving simple/efficient; the DB blob makes
backup and export trivial ("hand over one file" gets everything, even without the
files directory). Written once at upload; no edit in v1, so the two copies never
diverge. Acceptable cost because scale is tiny (two users, small files).
original_filename and mime_type kept as metadata for download/serving.
Backups remain JM's responsibility outside the app.

Auth integration

OIDC with PKCE against https://auth.jmopines.com.
Session cookie after successful callback.
/login, /callback, /logout, /healthz are public; everything else requires a valid session.
Group claim (hsa-users) gates access; otherwise 403.

Deployment

Runs as a systemd service in an LXC.
Caddy reverse proxy at https://hsa.jmopines.com (TBD: maisym.com vs jmopines.com).
SQLite DB + filesystem storage. No external dependencies (no Redis, no Postgres, no S3).

Operations

Soft delete supported (set deleted_at, hide from default list views).
No edit functionality in v1 — fix mistakes by deleting and re-adding.

Database export

Authenticated users (hsa-users group) can download the data for offline use.

Two endpoints:

GET /export/db — downloads a consistent copy of the SQLite database file. Because
  images are stored as BLOBs in the DB, this single file IS the complete dataset
  (metadata + all receipt images). This is the primary export.
  - Must NOT serve the live DB file directly (avoids locking/corruption against the
    running app). Use SQLite's online backup API (or VACUUM INTO a temp file) to
    produce a point-in-time snapshot, then stream that.
  - Content-Type: application/octet-stream; filename like hsa-export-YYYY-MM-DD.db.

GET /export/archive (optional convenience) — downloads a zip with the image files
  extracted to normal files (named by original_filename) plus a CSV/JSON of the
  metadata, for someone who wants the pictures as browseable files rather than
  inside a DB.
  - Streamed zip to avoid buffering large archives in memory.
  - filename like hsa-export-YYYY-MM-DD.zip.

Notes:
  - Amounts remain integer cents in the export; consumers divide by 100 for dollars.
  - Read-only operation; no app state is mutated.

Out of scope for v1

OCR / image content parsing
Reports, totals, dashboards
CSV / tax-software export
Reimbursement tracking (paid vs pending status)
In-place editing of existing receipts
Multi-tenancy or per-user data isolation
Notification / reminders

================================================================================
v2 — additions
================================================================================
These supersede the v1 "out of scope" entries for OCR-assisted entry (now present
via the classifier) and "Reports, totals, dashboards" (see Tally below). Same two
users, same auth model, same storage. Mobile-first still applies.

1. Skip-AI toggle on upload

A control on the upload form lets the user opt out of AI parsing for the current
receipt — for receipts they know are too hard to read, or to avoid spending an API
call on a bad result.
  - Default: AI parsing ON (auto-fill runs when a file is attached).
  - When "skip AI" is selected, NO /classify call is made; the user fills amount,
    date, category, and who by hand.
  - The model footnote stays visible but reads as disabled when skip is selected,
    so the user always knows whether a call will happen and which model it uses.
  - Skipping is per-upload, not a saved preference.

2. Duplicate-transaction warning before insert

Once a receipt has both a date and a dollar amount (whether typed or AI-filled),
check the DB for an already-posted receipt that looks like the same transaction,
and make the user confirm before inserting a possible duplicate.
  - Match: same receipt_date AND same amount_cents, among non-soft-deleted rows.
    When a "who" is set on both, prefer/highlight a same-person match; a match with
    a different or absent person is still shown as a weaker warning.
  - On a match, show the existing receipt's details: date, amount, category, who,
    uploaded_by, uploaded_at, original_filename (and a link to view it).
  - User chooses: Cancel (abort — nothing inserted) or Approve (insert anyway, as a
    deliberate duplicate). No new schema; this is a pre-insert read + confirm step.
  - No match → insert proceeds as today with no extra prompt.

3. Tally tab

A totals view (read-only). Bucketed by the YEAR of receipt_date.
  - Matrix: one row per person, one column per year that has data; each cell is the
    summed amount for that person in that year.
  - Include an "Unassigned" row for receipts with no "who".
  - Right margin column: grand total per year (all persons).
  - Bottom margin row: grand total per person across all years.
  - Bottom-right cell: overall grand total tracked.
  - Excludes soft-deleted rows. Amounts shown in dollars (cents / 100).

4. Recent uploads tab (by upload date)

A list of the most recently ADDED receipts, ordered by uploaded_at descending.
  - Show the 10 most recent, with a "Load next 10" control that pages further back
    (offset or cursor based).
  - Each row: receipt_date, amount, category, who, original_filename, and a link to
    view/download. Excludes soft-deleted rows.

5. Recent receipts tab (by receipt date)

Identical to #4 but ordered by receipt_date descending instead of uploaded_at —
"newest receipts" rather than "newest uploads". Same 10 + "Load next 10" paging.

6. People catalog integrity (bug)

The Manage page currently shows partial-name duplicates (e.g. both "Jude" and
"Jude Tremblay"). Only the canonical full names seeded from config.json
("First Last", per Person.Label) should exist as people.
  - Remove stray partial entries; keep only the config-seeded canonical labels.
  - Before deleting a partial entry, reassign any receipts that point at it to the
    matching canonical person so no receipt loses its "who".
  - Seeding must be idempotent: re-seeding from config.json must not create a second
    row for a person who already exists under the canonical label.

7. Per-parse cost shown in cents (¢)

After a receipt is classified, show the cost of that single AI call, in cents,
using the cent sign (e.g. "0.3¢"). The cost is computed locally from the API
response — no extra API call needed.
  - The Messages API response includes a usage object (input_tokens,
    output_tokens, and cache_creation/cache_read token counts). The classifier
    should capture these and return them alongside the suggestion.
  - Cost = input_tokens × input_price + output_tokens × output_price, using the
    active model's per-token rates. For the default model (Haiku 4.5): $1 per 1M
    input tokens and $5 per 1M output tokens — i.e. $0.000001/input-token and
    $0.000005/output-token. Cache-read tokens bill at ~0.1× input; treat them at
    the input rate unless we add exact cache pricing later.
  - Display in cents with the ¢ sign next to where the model footnote shows the
    model name, so the user sees both which model ran and what the scan cost.

Where the rates come from (this is the only maintenance cost of the feature):
  - Token counts are exact and free — they come straight from the response's
    usage object, no estimation.
  - Per-token PRICES are not available from any API (the Models API exposes
    capabilities, not dollars), so they live as hardcoded constants in a small
    rate table keyed by exact model id:
        haiku-4-5  → $1/1M in,  $5/1M out
        opus-4-8   → $5/1M in, $25/1M out
        sonnet-4-6 → $3/1M in, $15/1M out
  - This table is low-maintenance: Anthropic prices a specific model id once and
    ships price changes as NEW model ids, so an existing id's rate does not move.
    A new row is only needed when we adopt a new model — i.e. exactly when we'd be
    changing CLASSIFY_MODEL anyway.
  - Unknown model id (not in the table) → show the token counts but omit the ¢
    figure (or "cost: n/a"), never a guessed number. A stale table degrades
    gracefully instead of lying.

(Balance/spend indicator: dropped. The Anthropic API has no remaining-balance
endpoint, and a cumulative-spend readout was not wanted. Per-query cost above is
the only cost surface.)

9. Human-readable on-disk file layout

Replace the flat UUID filenames with a dated, amount-tagged layout under the
storage root, so the files directory is browsable on its own:

    <STORAGE_DIR>/<YYYY>/<MM>_<DD>_<dollars>.<cents>.<ext>

  - Year folder and MM/DD come from the RECEIPT date (not upload date); dollars
    and cents come from amount_cents (cents zero-padded to two digits, dollars not
    padded). Example: a $42.50 JPEG dated 2026-06-08 → 2026/06_08_42.50.jpeg.
  - The amount's decimal dot and the extension dot coexist fine — the extension is
    just the final dot-segment ("jpeg"); the stem is "06_08_42.50". (If that ever
    feels ambiguous, the accepted alternatives are an underscore "06_08_42_50" or
    bare cents "06_08_4250" — pick one and keep it consistent.)
  - All path components derive only from the date and amount (digits, underscores,
    one dot), never from the user-supplied original filename, so there is no path-
    traversal surface. original_filename stays as metadata in the DB.
  - Collisions (same date + amount + ext — legitimately possible since duplicates
    can be approved) get a numeric suffix on the STEM, starting at _1: the second
    file becomes "06_08_42.50_1.jpeg", the third "_2", etc. Use exclusive create
    (O_CREATE|O_EXCL) and increment the suffix on "already exists" so two concurrent
    uploads can't race onto the same name.
  - The chosen relative path is stored in receipts.file_path as today; the dual
    write still also stores the bytes as the DB blob, and serving continues to work
    from the blob regardless of the on-disk name.
  - Applies to NEW uploads only — existing UUID-named files keep their file_path;
    no backfill/rename of historical files in scope.
  - saveFile must create the year subdirectory (MkdirAll) before writing.

10. Additional attachments on a receipt

When uploading a receipt, the user can also attach one or more EXTRA files in the
SAME submission (a second page, an itemized list, an EOB, a photo from another
angle). Attachments are supplementary files for the same HSA expense — they are
not standalone receipts and carry no amount/date/category/who of their own; they
inherit the parent receipt's identity (including its date, which drives their
on-disk name). Adding attachments to an ALREADY-SAVED receipt is not supported yet
(no edit/detail page) — attachments are captured only at receipt-creation time.

Data model — new `attachments` table:
  - id (UUID)
  - receipt_id (FK → receipts.id; the parent expense)
  - uploaded_by (Authelia username/email)
  - uploaded_at (server timestamp)
  - file_path (relative path on disk — see layout below)
  - image_data (BLOB — the bytes, stored in the DB too, same as receipts)
  - file_size_bytes
  - original_filename
  - mime_type
  - deleted_at (nullable — soft delete)
  Index on receipt_id (and on deleted_at) for listing a receipt's live attachments.

Storage layout change — split receipts and attachments under the root:
  - Receipts move from <STORAGE_DIR>/<YYYY>/… to <STORAGE_DIR>/receipts/<YYYY>/…
    (the item-9 dated name is unchanged; only the "receipts/" prefix is added).
  - Attachments go to <STORAGE_DIR>/attachments/<YYYY>/…, named from the PARENT
    receipt's stem plus an attachment marker, e.g. a JPEG attached to the
    $42.50 receipt dated 2026-06-08 →
        attachments/2026/06_08_42.50_att.jpeg
    with the same exclusive-create disambiguation as receipts: the second
    attachment becomes _att_1, the third _att_2, etc. (Year/MM/DD/amount come from
    the parent receipt, so an expense's receipt and its attachments sort together.)
  - Same dual-write as receipts: bytes on disk AND as a DB blob, so the single .db
    export stays a complete dataset (receipts + attachments + images).
  - Existing receipts keep their stored file_path verbatim — file_path is the
    source of truth for serving location, so old (ROOT/<YYYY>/…) and new
    (ROOT/receipts/<YYYY>/…) paths coexist with no backfill. (In practice the DB
    was reset, so there are no legacy files to move.)

UI and endpoints:
  - The upload form gets an additional, OPTIONAL multi-file input ("Additional
    files", accept images + PDF, multiple). These ride along with the normal
    receipt submission to POST /upload — there is no separate attach endpoint.
  - Submission order: the receipt row is inserted first (so its id exists), then
    each attached file is saved and linked to it. Same MIME allowlist (images +
    PDF) and per-file size cap (MAX_UPLOAD_MB) as the receipt. No AI runs on
    attachments. AI auto-fill still reads only the primary receipt image.
  - The confirm page lists the saved receipt plus the count/filenames of any
    attachments.
  - GET /attachment/{id}/file serves an attachment's bytes from the DB blob (mirror
    of GET /receipt/{id}/file), so the recent list can link them.

Lifecycle / interactions:
  - Soft-deleting a receipt also hides its attachments (filter attachments by the
    parent's deleted_at, or soft-delete the children alongside the parent).
  - Attachments never affect Tally, duplicate detection, or the receipt counts —
    those operate on receipts only.
  - Export: attachments ride along automatically in the .db blob export; an archive
    /zip export (if/when added) lists them under their receipt.

Out of scope (for now):
  - Adding attachments to an already-saved receipt (would need an edit/detail page).
  - No AI parsing of attachments; no per-attachment metadata beyond the file.
  - No reordering UI beyond upload order.

11. Scheduled metadata-only DB backups

The app writes a periodic, metadata-only snapshot of the database to a local
directory so the receipt metadata always has a recent recoverable copy, without an
external cron job.
  - Built on the existing snapshot primitive (VACUUM INTO, then blob-strip): the
    backup excludes image blobs from BOTH receipts and attachments, so it is tiny.
    The on-disk files and the full-blob export (/export/db) remain the source for
    the images themselves.
  - On by default, weekly. Config: BACKUP_DIR (default ./data/dbbackup),
    BACKUP_INTERVAL (Go duration sets the frequency, default 168h; set 0 to
    disable), BACKUP_KEEP (most-recent copies to retain, default 8).
  - Files are named hsa_sqlite_backup_<YYYY_MM_DD>.db — one per day, sorting
    chronologically; older copies beyond BACKUP_KEEP are pruned. A same-day re-run
    is a no-op (the dated file already exists).
  - Restart-safe: on startup it backs up immediately only if the newest existing
    backup is older than the interval, so frequent restarts don't spam the dir and
    a long gap is covered right away. Runs in a background goroutine for the life
    of the process; failures are logged, never fatal.

Storage-root note: STORAGE_DIR is the storage ROOT (not the receipts subdir).
Receipts live under <STORAGE_DIR>/receipts/<year>/ and attachments under
<STORAGE_DIR>/attachments/<year>/ (default STORAGE_DIR=./data).

12. Crash recovery and durability at boot

The database must survive an app crash or kill mid-write without corruption or a
half-written row, and recover automatically on the next boot — no manual repair.
  - SQLite runs in WAL mode (PRAGMA journal_mode=WAL, set on open). Every write is
    an atomic transaction, so an interrupted write is either fully applied or not at
    all. Opening the file is self-healing: SQLite rolls forward committed
    transactions from the WAL and discards any incomplete tail, bringing the DB up
    at its last committed state. There is no custom "revert" logic — SQLite's own
    recovery is the mechanism, and reimplementing it would be less safe.
  - On open the app runs PRAGMA quick_check to confirm the recovered file is sound.
    A crash-interrupted write passes (WAL recovery already healed it). A failure
    means genuine corruption (e.g. disk failure); the app refuses to start with a
    message pointing at BACKUP_DIR, so the operator restores rather than running on
    a corrupt DB.
  - Dual-write caveat: a receipt's (or attachment's) file is written to disk BEFORE
    its DB row is inserted, so a crash in between can leave an orphan file with no
    row. This is harmless — extra bytes on disk, never a row missing its data — and
    is not auto-cleaned. The DB/blob is the source of truth for serving.
  - Recovery from true corruption (beyond crash-consistency) is restore-from-backup:
    the metadata-only daily backup (item 11) recovers the records; the full .db
    export (/export/db, blobs included) recovers records + images.