dompdf does not use a declared line-height directly. It scales it by the font's
own height:
rendered = declared x (ascent + descent) / unitsPerEm x font_height_ratio
The bundled Noto Sans reports 1.362 for that middle term, so at dompdf's stock
font_height_ratio of 1.1 every line-height in every document came out 1.4985x
what the CSS asked for. Chromium honours the declared value exactly. That one
factor was the whole vertical disagreement between the two drivers.
Setting the ratio to 1/1.362 cancels the font term. Measured on a declared 15px
(11.25pt): 16.86pt at the stock 1.1, 15.32pt at 1.0, and 11.25pt at this value --
identical to Chromium.
Across the seven document templates the worst-edge ink difference falls from
roughly 70-150pt to under 30pt on six of them, and to 3.4pt on invoice1. The
exception is estimate1, which moves the other way: with the line-height noise
gone, a float and padding difference in its address block is now the dominant
term there. That is a separate problem this exposes rather than causes.
Worth recording that an earlier compensation shim had arrived at 1.5 empirically
and was right: 1.1 x 1.362 = 1.4985. I argued against it on the strength of a
test that used font-family: sans-serif, which resolves to a built-in core font
and so never exercised the embedded Noto Sans path where the scaling happens.
The measurement was wrong, not the constant.
PdfLineHeightTest pins the invariant -- the font's reported height equals the
font size, so a declared length renders at that length -- and needs no Gotenberg,
so CI holds it. Swapping the default face or taking a dompdf upgrade that changes
the computation now fails a test rather than quietly reintroducing the drift.
Claude-Session: https://claude.ai/code/session_01QmECndmNZwzN65Zz9P87dF
* fix(pdf): stop an unresolvable template taking the PDF route down
RealisticDemoSeeder::seedEstimate() never set template_name, while seedInvoice()
has always set invoice1. Every demo estimate therefore had '', so
findFormattedTemplate() returned null and EstimateService did $template['custom']
on it -- a 500 on the estimate PDF route, on either driver, since the exception
is thrown before a driver is reached. That is the "Unable to load document
preview" people were seeing.
The seeder now sets estimate1, but seeding was only how this surfaced. The stored
name is validated when a document is saved through the UI and nowhere else:
seeders, imports, recurring-invoice copies and rows predating that validation all
bypass it, and a template can also be deleted from disk after the fact. A name
that cannot be resolved should fall back to the default design, not take the
route down.
PdfTemplateUtils::resolveView() -- already the resolver for payment receipts and
reports -- gains an optional fallback and tries each candidate as custom then
built-in. Both document services collapse to a single call and can no longer
index null. The fallback logs a warning, so a bad name stays visible rather than
being silently swapped.
Also casts two nulls in GeneratesPdfTrait: an address line or custom field that
was never filled in reaches htmlspecialchars() and strtr() as null, which every
PDF render was emitting a deprecation for on PHP 8.4 and would be an error on 9.
Claude-Session: https://claude.ai/code/session_01QmECndmNZwzN65Zz9P87dF
* feat(pdf): a command that measures the two drivers against each other
"The PDF looks different" has been diagnosed by eye every time, because nothing
compares the renderers. Asserting on PDF bytes is useless and rendering through
Gotenberg needs a live service, so the suite has never covered it.
pdf:compare renders each stock template through both drivers and reports the
page box, page count and the bounding box of the text on page one, then flags
any template whose ink lands more than --tolerance points apart. It goes through
the real document services, so it exercises the same shared view data and
template resolution a request would.
Two things it has to get right to be honest:
Comparing designs means persisting the template choice -- InvoiceService reads it
back with Invoice::find($id)->template_name, so assigning in memory silently
compares the same design every row. The run happens inside a transaction that is
always rolled back.
Page numbers are turned off for the duration. They are a Chromium capability with
no dompdf equivalent, so leaving them on puts ink at the foot of every Gotenberg
page and drowns out every difference worth seeing -- which is exactly what the
first run of this command did.
Word positions come from poppler's pdftotext, which is on most dev machines but
not in the app container; without it the command still compares page geometry and
says what it could not check.
Claude-Session: https://claude.ai/code/session_01QmECndmNZwzN65Zz9P87dF
* fix(pdf): let the two renderers agree on the items table, and drop the shim
Two parts: the stock templates stop doing their own page margins, and the items
table stops relying on a property that does not apply to it.
Page margins. The templates carried their own via `html { margin-top: 50px }`,
which predates page setup owning them. dompdf largely collapses that margin;
Chromium honours it and adds it to the page box, so the same template came out
38px from the top on one renderer and 77px on the other. The html rule is gone
and body is reset instead, which is what makes the page box agree. Headers that
were positioned absolutely at a negative offset -- only possible because of that
margin -- are back in flow.
The items table. Every stock template sets `table { border-collapse: collapse }`,
and CSS says padding does not apply to a table in that mode. dompdf applies it
anyway; Chromium follows the spec and drops it, so the table's `padding: 0 30px`
inset the content on one renderer and not the other. Measured in isolation: with
border-collapse, content starts at x=24.0 on dompdf and x=1.5 on Chromium -- the
full 30px. All of the table's spacing moves to .items-table-wrapper, a plain
block both engines treat the same, using padding so nothing collapses through it
either.
Measured across the seven document templates, that closes the horizontal gap
outright: xMin was 57 on dompdf against 37 on Chromium for five of them, and is
now within 3pt on all seven.
GotenbergStockTemplateCompatibility is removed. Its premise was that dompdf
inflates declared line heights by 1.5x, and that does not hold: rendering the
same text at 12px, 18px, 36px and unitless 1.0/1.5 through both engines gives
line spacing within 0.5pt every time. It also applied its multiplier to the
reports, where line-height 21px pairs with font sizes of 14, 16 and 20px -- so
.report-footer-value at a 1.05 ratio was being blown out to 31.5px, half again
taller than dompdf renders it.
A residual vertical difference remains and is localised, not guessed at: it
accumulates only in the address blocks, which are <br>-joined text emitted by
getFormattedString() with an <h3> in front. Reduced to that construct alone,
Chromium steps 11.25pt per line -- exactly the declared line-height: 15px --
while dompdf steps 14.4pt. That needs deciding on its own terms rather than a
global multiplier, so it is left visible and measurable via pdf:compare.
Claude-Session: https://claude.ai/code/session_01QmECndmNZwzN65Zz9P87dF
* fix(pdf): restore the stock template design
Two regressions from the page-setup work, both visible on the page.
The coloured header band stopped bleeding to the paper edges. invoice2 and
estimate2 are built around a full-width band, and it now sits in normal flow at
the top of body, so it only reaches the edge when the page margin is nothing.
#728 defaulted margins to 1.2cm on the reasoning that it matched dompdf's
built-in default and so kept existing output unchanged. That was the wrong
reference: the templates are drawn for a zero margin and carry their own 30px
insets, and Gotenberg rendered them at zero before #728, which is the intended
look. Margins now default to nothing. Setting one still works and is honoured by
both drivers, at the cost of the band no longer reaching the edge.
A bare `0` is valid CSS and the only length needing no unit, so CssLength and
PdfPageSetup accept it -- without that the new default would have thrown on
every render.
The totals block was pushed in from the items table's right edge. Fixing the
border-collapse padding problem moved the table's 30px inset onto a wrapper that
contains the whole partial, so it stacked on the insets the hr (25px) and the
totals container (25px) already had. Those two were always honoured by both
renderers; only the table's own padding was not. The inset now lives on a div
wrapping just the table, and the wrapper keeps vertical spacing only, which
restores the original 30px/25px relationship rather than inventing a new one.
Also drops the negative margin-bottom that pulled the addresses up into the band
and hid "Bill to,", and removes a stray `bottom: 0px` on invoice1's
.header-bottom-divider that combined with `top: 90px` to stretch the rule down
the page.
Checked by rendering, not only by measurement: invoice2 and invoice1 on both
drivers now match the intended design. pdf:compare puts the two renderers within
a few points horizontally on all seven documents, xMin 21-22 and xMax 564-575.
The remaining vertical difference is the address-block line spacing documented
earlier and is unchanged by this.
Claude-Session: https://claude.ai/code/session_01QmECndmNZwzN65Zz9P87dF
* feat(demo): make the demo data actually demo the product
The demo company had no address row, and Invoice::getCompanyAddress() returns
false outright in that case, so every seeded document rendered with an empty
company block -- the name only appeared because the header falls back to it when
there is no logo. Several headline features had no demo data at all: zero tax
types, zero notes, zero recurring invoices, zero custom fields.
DemoSeeder, which the test suite and reset:app both run, now creates Acme Inc
with a postal address, tax ids and a country -- the fields the default address
format actually renders. The address is created through the relation, as
CompaniesController does, so company_id is set and type/user_id/customer_id stay
null: Company::address() is an unscoped hasOne, so anything else carrying that
company_id would be picked up as the company's own.
It also stops trusting currency id 1. Migration 2025_08_18 inserts Algerian
Dinar via firstOrCreate() before any seeder runs, so on a fresh migrate+seed the
demo priced everything in "DA". RealisticDemoSeeder already worked around this
for itself; resolving USD by code fixes it at source for reset:app and the tests
too.
RealisticDemoSeeder gains a logo, two tax types, a notes library, custom fields
and an active recurring invoice. Notes are seeded twice over on purpose: the
library and a document's notes column are unrelated in this application -- there
is no foreign key, and is_default only drives a badge in the settings list, so
nothing pre-fills a document with one.
Tax is applied at document level to most but not all documents, so the demo has
a zero-rated example in it. The arithmetic is the caller's: the service layer
trusts whatever amount it is handed rather than recomputing it, so tax is
rounded once off the subtotal and carried through total, due_amount and every
base_* twin -- miss due_amount and a paid invoice renders as part-paid.
Custom fields are on Customer, the only model_type with a create/edit UI end to
end. The PDF renders only model_type 'Item', which would add a column to the
items table and disturb a layout that was just squared up across both drivers.
The logo is a generated Acme mark rather than one of InvoiceShelf's own, which
would read as InvoiceShelf billing the customer.
Also documents both seeders in AGENTS.md. RealisticDemoSeeder was referenced
nowhere outside database/seeders/, which is a poor place to keep the thing that
makes the app look real.
Claude-Session: https://claude.ai/code/session_01QmECndmNZwzN65Zz9P87dF
* fix(pdf): tighten the spacing the old absolute header left behind
invoice2 and estimate2 carried three stacked top offsets -- content-wrapper's
60px margin plus address-container's 18px margin and 20px padding -- 98px of
dead white between the coloured band and the first line of content. They existed
because the band used to be position: absolute and out of flow, so everything
below had to be pushed clear of where it visually sat. The band takes its own
height now, so the compensation is just a gap. Collapsed to a single 32px.
Only those two templates had it, which is the tell: they are exactly the two
whose headers were absolutely positioned.
Also pins the margins on the <h3> the address formats emit. Left to the
user-agent default it pushed the company column out of line with the Bill to /
Ship to columns beside it, so the three column headings started at three
different heights. They line up now.
That h3 is also where the two renderers were measured drifting apart, and
pinning it narrows invoice2 from 85.8pt to 72.0pt and estimate2 from 84.4 to
76.8. The templates without a coloured band barely move, which places the rest
of the difference in the per-line spacing of the <br>-joined address lines
rather than in the heading -- consistent with the isolated measurement earlier
(dompdf 14.4pt per line against Chromium's 11.25pt) and still open.
Claude-Session: https://claude.ai/code/session_01QmECndmNZwzN65Zz9P87dF
FontService writes absolute host paths into the @font-face rules:
src: url("/var/www/html/storage/fonts/NotoSansSC-Regular.ttf")
dompdf shares that filesystem so they resolve. Chromium runs inside the Gotenberg
container and cannot see any of it, so every installed font package silently
failed to load and documents fell back to whatever fonts that image happens to
ship. The docs recommend Gotenberg specifically for mixed-script documents, which
made this exactly the wrong way round -- it worked only by accident, because
Chromium's own font set covers more than dompdf's single-font behaviour.
The font files now travel with the document as Gotenberg assets, and the rules
are rewritten to name them. Gotenberg unpacks assets next to index.html, so a
bare filename resolves.
Only fonts the markup actually references are sent. A CJK package is several
megabytes and has no business riding along on a request that never mentions it.
Confirmed against a stock gotenberg:8, reading the fonts back out of the
rendered PDF:
before AAAAAA+LiberationSerif (Gotenberg's fallback)
after AAAAAA+NotoSans-Regular (the app's own font)
Claude-Session: https://claude.ai/code/session_01QmECndmNZwzN65Zz9P87dF
Generated files carried no document properties at all, so an archive of them
showed a column of blank titles and no author. Title, Subject, Author and
Creator are now written from the document number and company, on both drivers:
dompdf via addInfo(), Gotenberg via metadata().
dompdf needed more than the API call. It reads Title from the <title> element
during render(), which happens after addInfo(), so metadata set through the API
alone was silently overwritten by whatever the template put there and the two
drivers disagreed about what the file was called. The title is written into the
markup as well, escaped.
Also adds an archival format setting for Gotenberg: off, PDF/A-1b, -2b or -3b.
PDF/A-3 is what the EU e-invoicing formats expect. Verified against a stock
gotenberg:8 -- LibreOffice inside the image does the conversion and the output
carries the right pdfaid:part in its XMP -- so no extra components are needed.
A fixed list rather than free text, because the SDK forwards whatever it is
given and an unsupported value would surface only as an HTTP error from the
service at render time. Empty is a real choice meaning an ordinary PDF, so it
overrides an env default rather than falling through it.
Gotenberg only: dompdf cannot produce PDF/A.
Claude-Session: https://claude.ai/code/session_01QmECndmNZwzN65Zz9P87dF
Takes over #690 by csoscd. The companion-view idea is theirs; this reworks it
onto the shared page setup and fills in the gaps that stopped it landing.
A `{template}_header` or `{template}_footer` view next to a template is rendered
alongside it and repeated by Chromium on every page. The suffix resolves through
the pdf_templates:: namespace too, so custom templates get it with no extra
wiring.
Two things had to change for that to be useful.
Companion views are now hidden from the template picker. getFormattedTemplates()
lists every .blade.php it finds, so an invoice1_footer would otherwise appear as
a separately selectable template with no preview image -- the feature would have
introduced that the moment anyone used it.
And it does something out of the box. #690 shipped no companion views, so both
of its margin settings were visible no-ops until someone hand-wrote a Blade file.
Instead there is a pdf_page_numbers setting, off by default, that supplies a
footer when a template has none. A template's own companion still wins, so
turning page numbers on cannot overwrite a designed footer.
The setting sits under Gotenberg because only Chromium can repeat a footer;
dompdf has no equivalent. Its value is still carried by the dompdf form so
saving from there cannot clear the choice -- the field is absent from that
payload, and the controller only writes it when present.
Margins come from the page setup rather than #690's separate header_margin and
footer_margin. Chromium draws header and footer inside the page margin, so the
existing margins are the space they occupy; two more settings for the same
distance would have been a second way to say the same thing.
Verified against a live gotenberg:8 on a two-page document: off produces no
footer, on produces "1/2" and "2/2" on the respective pages, and a companion
footer replaces both.
#690's own test is not carried over. It asserted nothing: a bare
View::shouldReceive('exists') is an allowance rather than an expectation,
andReturn(false) never entered the companion branch, and the call sat inside
try { } catch (Throwable) { }, so it passed with the feature deleted.
Claude-Session: https://claude.ai/code/session_01QmECndmNZwzN65Zz9P87dF
Paper size was a Gotenberg-only setting stored as a single "210mm 297mm"
string. dompdf had no page settings at all: size was pinned to config/dompdf.php's
fixed 'a4', its top-level `orientation` key was read by nothing (the installed
barryvdh v3 builds options only from `defines`), and margins were whatever
dompdf's own stylesheet said. So the two drivers disagreed about margins by
default -- dompdf 1.2cm, Gotenberg hardcoded to zero -- and selecting dompdf
silently discarded the paper size.
Replaces gotenberg_papersize with pdf_paper_width / pdf_paper_height /
pdf_orientation / pdf_margin_{top,right,bottom,left}, saved and applied for
either driver. Width and height are separate CSS lengths because that is the
only lossless shared notation: Gotenberg has no named sizes, and dompdf's named
table cannot express everything Gotenberg accepts. Named presets (A3/A4/A5/
Letter/Legal) are a convenience in the UI that resolve to a pair of lengths.
PdfPageSetup resolves it once and translates: a points array plus an orientation
argument for dompdf, CSS lengths plus landscape() for Gotenberg. Both are handed
the portrait pair, since each swaps the axes itself. Gotenberg's margins() takes
top, bottom, left, right, which is not the CSS order.
dompdf exposes no margin API, so DompdfDriver injects an @page rule -- at the
top of <head>, so a template declaring its own still wins. Doing it in the driver
rather than a Blade partial means custom templates get it without including
anything.
Margins default to 1.2cm, dompdf's existing default, so Gotenberg starts
matching it rather than rendering edge-to-edge. Verified against a live
gotenberg:8: A4 portrait, A4 landscape and Letter at zero margins all come out
with the same page box and the same ink offsets on both drivers.
A malformed length now throws rather than being ignored. Blank still falls back,
but a value that is set and wrong is an operator mistake, and the drivers would
otherwise fail differently: dompdf throws converting to points, Gotenberg would
forward the string and render at some other size.
Also here:
- Migration splits an existing gotenberg_papersize into the new pair. It earns
its place because that key ships in 2.x, not just a 3.x alpha, so a stable
install that chose Letter would otherwise come back up on A4. Drops
gotenberg_margins, which 2.x also stores and neither driver ever read.
- Removes EnvironmentManager::savePDFVariables/getPDFConfiguration, which had no
caller anywhere, and the unused EnvironmentManager injection in the controller.
- config/dompdf.php: drops the dead `orientation` key and defaults enable_remote
to false, matching .env.example, which sets it explicitly and explains why.
Installs predating that line were falling back to true.
- Retires the settings.pdf.footer_text and pdf_layout strings, which no component
referenced.
Claude-Session: https://claude.ai/code/session_01QmECndmNZwzN65Zz9P87dF
PdfDriver and ResponseStream existed but nothing implemented them. The factory
returned the vendor dompdf wrapper for one driver and a bespoke class for the
other, so the two were never held to the same shape. Three things had slipped
through that gap.
Report PDFs answered 403 for everyone. The five report routes carry no company
header, so ScopeBouncer is not in their middleware stack and the ability scope
was never set; 'view-financial-reports' is stored scoped to a company, so the
check could not pass. They now scope to the company named in the URL. The policy
still checks membership, so this grants nothing new. Also firstOrFail() on the
hash lookup, so an unknown company is a 404 rather than a 500 on a null.
Report downloads were fatal on Gotenberg. GotenbergPdfResponse had no download(),
and the report controllers are its only callers. Added, alongside stream() and
output(), with the whole set now on the interface.
Streamed documents carried an HTTP preamble. GeneratesPdfTrait wrapped
$pdf->stream() -- already a Response -- in another response()->make(), which
stringified it and prepended "HTTP/1.0 200 OK" plus headers to the file. Readers
scan the first kilobyte for %PDF so nobody noticed, but the bytes were malformed.
Passing ->output() fixes it, and the render test now asserts the position.
Two driver-parity settings, both checked against a real gotenberg:8 rather than
inferred: emulateScreenMediaType(), because Chromium defaults to print media
while config/dompdf.php renders as screen, so a @media print rule applied on one
driver and not the other; and printBackground(), which turns out to affect only
the root background, since Chromium paints element backgrounds either way. No
stock template sets a body background, so that one changes nothing today and is
here to keep custom templates consistent across drivers.
Claude-Session: https://claude.ai/code/session_01QmECndmNZwzN65Zz9P87dF
* fix: allow Gotenberg to reach private/Docker-internal hosts (Issue #688)
The SSRF guard introduced in #664/#671 correctly blocks arbitrary
private URLs, but also prevents legitimate use-cases where Gotenberg
runs alongside InvoiceShelf in a Docker Compose network (e.g. the
default http://pdf:3000 service name resolves to a private IP).
Add a `gotenberg_allow_private_host` setting (env:
GOTENBERG_ALLOW_PRIVATE_HOST, default false) that:
- skips PrivateNetworkGuard in GotenbergPdfDriver
- skips PublicHttpUrl validation in PDFConfigurationRequest
- exposes a clearly-warned toggle in the admin PDF settings UI
- is persisted to the settings table and loaded via AppConfigProvider
A disabled guard is safe for controlled private networks (Docker
Compose, LAN); it must never be enabled for untrusted hosts. The UI
surfaces a prominent warning to communicate this constraint.
Closes#688
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* refactor(gotenberg): scope the private-host exemption to a declared host
Reshapes the escape hatch from a boolean admin setting into an
environment-declared host allowlist.
The driver streams the upstream response body back as the PDF, so a
mis-set Gotenberg host is full-response SSRF — pointed at a link-local
metadata endpoint it returns cloud credentials. A blanket "allow
private" switch left that reachable: gotenberg_host stays editable from
the admin UI, so any install that enabled the switch to run a sidecar
could have the host repointed at an internal service. The population the
flag existed to serve was exactly the population it failed to protect.
GOTENBERG_ALLOWED_PRIVATE_HOST now names the single host that may skip
the guard. Only that exact value is exempt; every other private target
stays blocked. GotenbergHostPolicy owns the comparison so the save-time
rule and the runtime driver guard cannot drift, and normalises case,
trailing slash and surrounding whitespace on both sides.
Being env-only also drops the settings-table key, the AppConfigProvider
branch and the whole admin UI surface — the toggle there could not be
switched on in any case, since BaseSwitchSection has no slot and was
passed no v-model, so the child BaseSwitch was discarded and the value
never changed from false.
Restores the gotenberg_margins validation rule, which the previous
revision replaced rather than added alongside.
Tests cover both directions, including that declaring one private host
does not exempt another; sabotaging the policy to always exempt fails 16
of the 22.
Co-authored-by: csoscd <csoscd@users.noreply.github.com>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: Darko Gjorgjijoski <dg@darkog.com>
Co-authored-by: csoscd <csoscd@users.noreply.github.com>
Ports #639 to 3.x — the original targets the now feature-frozen 2.x line.
Address::country_name now resolves the localized country name for the
current app locale via Symfony\Component\Intl\Countries, falling back to
the stored name on lookup failure. Adds symfony/intl + a unit test.
Co-authored-by: Lukas Selch <selchlukas@icloud.com>
v3 port. Invoice/estimate/recurring creation and update accepted total,
sub_total, tax and due_amount straight from the request with no recalculation,
letting a client persist financial totals that don't match the line items
(and corrupt the invoice update due-amount/paid-amount logic which keyed off
the client total).
- Adds App\Support\DocumentTotals (mirrors the front-end calc) trusting only
price/quantity/discounts/tax-line amounts.
- getInvoicePayload/getEstimatePayload/getRecurringInvoicePayload override the
client totals; the shared DocumentItemService::createItems recomputes each
item total; InvoiceService::update keys its due-amount logic off the
recomputed total.
Adds DocumentTotals unit tests + a feature test proving a tampered invoice
total is ignored; existing create/update tests no longer assert the now
server-authoritative derived totals.
v3 port. The Gotenberg PDF driver was missed when the SSRF guards were added
to the AI, exchange-rate and file-disk drivers: gotenberg_host was validated
only with 'url', and the driver POSTs the rendered HTML to it.
Reuses the existing infrastructure (consistency with the other drivers):
- Wires App\Rules\PublicHttpUrl into the gotenberg_host validation rule.
- Adds PrivateNetworkGuard::assertAllowed() in GotenbergPdfDriver before the
outbound call (covers env/seed/stale config + DNS rebinding).
Adds a unit test asserting the gotenberg_host rule rejects private/loopback/
link-local addresses and allows a public one.
v3 port. orderByField/orderBy were passed straight into Eloquent's orderBy()
in every model's scopeWhereOrder (and Invoice::scopeApplyFilters), allowing
arbitrary SQL in the ORDER BY clause.
Adds App\Support\SafeOrderBy::apply() (plain/table-qualified column identifier
only, asc/desc clamp) and routes all 10 model sort sinks through it. Aliased
sorts (e.g. estimates by customers.name) stay valid.
Adds unit tests for injection rejection, plain + aliased columns, direction clamp.
The AI chat assistant scoped tool queries by company but ignored the
per-user Bouncer abilities the rest of the app enforces, so any `use ai`
holder could read customers, invoices, payments, and company financials
their role couldn't otherwise see. Each AiTool now declares a required
ability (entity-aligned); the registry hides unauthorized tools from the
model and refuses to execute them as a backstop.
Separately, admin/owner-supplied URLs were fetched server-side with no
guard against private/reserved targets (SSRF): the AI base URL, the
CurrencyConverter "DEDICATED" exchange-rate URL, and S3/Spaces file-disk
endpoints. A shared PrivateNetworkGuard now backs a PublicHttpUrl
validation rule (save-time) and runtime guards in each driver.
- AiTool::requiredAbility() + mapping across all 12 tools
- AiToolRegistry filters schemas() by ability and re-checks in execute()
- PrivateNetworkGuard / BlockedUrlException / PublicHttpUrl rule (new)
- Rule wired into AI config (service + 3 controllers), exchange-rate,
and file-disk endpoints; runtime guards in OpenRouterDriver,
CurrencyConverterDriver, and FileDiskService
- Tests for ability filtering, the guard, the rule, and 422 rejections
Extracts the two inline LLM prompts (AiAssistantService's chat system
prompt and AiTextGenerationService's writing preamble) out of PHP
heredocs and into plain-markdown template files under resources/ai/
prompts/. Each file can now be edited without opening a service
class, without wrestling with PHP string interpolation, and with
proper markdown syntax highlighting in editors.
A tiny PromptLoader helper at app/Support/Ai/PromptLoader.php reads
the file and does {{placeholder}} substitution via strtr() — no
Blade, because Blade's {{ $var }} HTML-escapes ampersands and quotes,
which is wrong for LLM prompts (a company called "Smith & Co" would
be sent as "Smith & Co"). Missing templates throw RuntimeException
so they fail loud during development.
Pure refactor: no prompt wording changes. Existing AI feature tests
(AiChatFlowTest, AiGenerationTest) pass unchanged — they assert on
message structure via ScriptedAiDriver, not on prompt content. Three
new unit tests in PromptLoaderTest lock the helper's contract:
placeholder substitution, no-var loading, missing-file error.
Second phase of the AI feature. Users can now open a slide-in chat drawer from the SiteHeader and ask natural-language questions about their company's invoices, customers, payments, and expenses. The LLM invokes pre-defined read-only tool functions (scoped to the current company at execute time) to fetch data and synthesize answers.
**Database** — new ai_conversations and ai_messages tables. Messages are stored in OpenAI's chat format so AiAssistantService serializes a conversation into an API request with zero translation. Columns: role, content, tool_call_id, tool_calls JSON, model, tokens_in, tokens_out. Conversations are scoped (company_id, user_id) — one user's chats are invisible to everyone else, even inside the same company. Foreign-key cascade deletes.
**Tool infrastructure** — AiTool abstract base + AiToolRegistry singleton (registered in a new AiServiceProvider). The base class enforces the security rule: every tool's execute() receives companyId and userId as injected parameters; tools' JSON schemas NEVER include a company_id field. An LLM physically cannot pass a company_id and escape tenancy. Modules can register their own tools by resolving the registry from their own ServiceProvider::boot().
**Nine built-in tools**: search_invoices, get_invoice, search_customers, get_customer, list_recent_payments, list_overdue_invoices, get_company_stats (aggregates for named periods), search_items, list_expense_categories. All read-only; no mutations. Each returns JSON-encodable data the LLM can parse.
**AiAssistantService orchestration loop** — the heart of Phase 2. Flow: persist user message → build payload from system prompt + recent history (40-message window) + new user message → call driver.chatCompletion with tools → if tool_calls, execute each one via the registry (with injected scope), persist tool result, loop → if plain text, persist and return. Hard cap at 5 iterations to prevent runaway LLMs. System prompt pins the assistant to this company's data and forbids mutation.
**Controllers + policy + rate limit** — POST /api/v1/ai/chat runs the orchestration loop. GET/PATCH/DELETE /api/v1/ai/conversations for CRUD. AiConversationPolicy enforces user_id+company_id match on every action. A new 'ai' RateLimiter in RouteServiceProvider throttles to 30 req/min per (user, company). New 'use ai' Gate defined in AppServiceProvider returns true for any authenticated user — the per-company kill-switch still goes through AiConfigurationService::resolveForCompany.
**Frontend** — new features/company/ai/ folder with a Pinia store (ai-chat.store.ts) holding drawer state, current conversation, messages, and loading flags. AiChatDrawer.vue is a slide-in panel teleported to <body>, mounted globally in CompanyLayout.vue when bootstrap reports ai.enabled && ai.chat_enabled. Sub-components: AiChatMessage (user bubbles vs assistant bubbles), AiChatMessageInput (Enter submits, Shift+Enter newline), AiChatConversationList (sidebar with 'new chat' button, rename, delete). A SparklesIcon button in SiteHeader toggles the drawer.
**Driver test double** — tests use a ScriptedAiDriver registered via AiDriverFactory::register('scripted', ...) that returns pre-queued AiChatResponse objects. Feature tests cover: happy path (new conversation + message persistence), tool-call loop (multi-round-trip with search_invoices), runaway-loop cap, driver-throws path, ai_enabled=NO rejection, chat role disabled rejection, per-user conversation visibility, cross-user policy enforcement, cascade delete.
388 tests pass (was 372, +16 new). Pint clean. npm run build clean. Phase 3 (WYSIWYG text generation popup) is the remaining follow-up.
Foundation for the AI chatbot + text generation feature. Phase 1 is infrastructure only: driver plumbing, configuration storage with encrypted API keys, global vs per-company resolution, admin + company UI pages, and an optional installer wizard step. The chat assistant and text-generation WYSIWYG integration come in later phases.
**Driver plumbing (app/Support/Ai/)** — AiDriver abstract, AiDriverFactory, AiException, AiChatResponse DTO, OpenRouterDriver concrete implementation. OpenRouter is the OpenAI-compatible aggregator that unlocks hundreds of models behind one API key and one request shape — ideal as the default v1 driver. Drivers are extensible the same way exchange rate drivers are: the module Registry's generic registerDriver('ai', ...) machinery plus a typed Registry::registerAiDriver() convenience wrapper (shipped in the upstream invoiceshelf/modules package in a paired commit).
**AiConfigurationService** — mirrors MailConfigurationService shape but with one deliberate deviation: API keys are encrypted at the service layer via Crypt::encryptString before persistence. OpenRouter bearer tokens have much bigger blast radius than SMTP passwords. Same settings / company_settings tables, same global-vs-per-company pattern, same use_custom_ai_config override toggle. Resolution order: global ai_enabled must be YES, then the company either overrides via use_custom_ai_config=YES (and can opt out with ai_enabled=NO inside the override) or inherits the global config.
**Controllers** — Admin/Settings/AiConfigurationController (global CRUD + driver list + test connection), Company/Settings/CompanyAiConfigurationController (per-company override + test), Setup/AiConfigurationController (installer wizard step, skippable with explicit ai_enabled=NO). API key is always masked as '********' in GET responses — the frontend submits the placeholder back on save and the backend preserves the stored value.
**Installer wizard** — new optional step 7 'AI' between Mail and Account. Default OFF with a Skip button. MailView.vue now routes to installation.ai instead of installation.account; installation.ai then routes to installation.account. Step order comment updated in routes.ts.
**Admin + Company settings pages** — AdminAiConfigView (no toggle, always global) and AiConfigView (with use_custom_ai_config BaseSwitchSection that auto-saves OFF). Both share AiConfigurationForm which renders the driver selector, API key input with show/hide, driver-specific config_fields (base_url for OpenRouter), and per-role enable toggles with free-text model inputs backed by a datalist of suggested models from driver metadata.
**Bootstrap endpoint** — adds an ai block to the response: { enabled, chat_enabled, text_generation_enabled }. All three are booleans resolved through AiConfigurationService::resolveForCompany(). Never leaks the API key. Frontend feature flags read from bootstrapData.ai to decide whether to show Phase 2/3 UI.
**Bouncer ability** 'manage ai config' added to SettingsPolicy, gated on isSuperAdmin() (same pattern as manage email config, manage pdf config).
**Tests** (22 new) — Unit: AiDriverFactory resolves built-in + Registry-contributed drivers, rejects unknown, merges availableDrivers. AiConfigurationService: encryption round-trip, resolution order (3 cases: global off, inherit global, override with company key, override with opt-out), makeDriver null/instance cases, listDrivers metadata. Feature: admin save + read with api key masking, preserve-on-placeholder behavior, company toggle ON/OFF semantics, bootstrap ai flags reflect resolution, company opt-out path.
372 tests pass (was 350, +22). Pint clean. npm run build clean. Phase 2 (chat assistant + tool calling) and Phase 3 (WYSIWYG text generation popup) are separate follow-up commits — this one is the foundation only.
Two plural outliers were the only directories in Services/ and Support/ that didn't follow the singular naming convention we normalized in commit 947d00a9 (Documents -> Document).
**Integrations/ExchangeRate/ -> ExchangeRate/.** The Integrations/ umbrella was both plural AND introduced an extra layer of nesting no other Support subdir had (Pdf/, Hashids/, Module/, Update/ are all one level deep). Dropping the umbrella fixes both problems in one move and matches the existing shape. When AI providers eventually land, they follow the same pattern as Support/Pdf/: a sibling subdir at Support/Ai/, not buried under an umbrella.
**Formatters/ -> Formatting/.** Formatter (singular noun) would have been awkward; Formatting (gerund describing the capability) reads naturally as a namespace segment. The subdir holds DateFormatter, TimeFormatter, TimeZones — classes whose common thread is 'things that do formatting', which the gerund captures better than either the singular or plural noun form.
10 files renamed, 4 consumers updated (DriverRegistryProvider, ExchangeRateProviderService, FormatsController, ExchangeRateDriverFactoryTest). 350 tests pass, Pint clean.
Services/Integrations/ExchangeRate/ and Services/Pdf/ were both mostly Support-shaped: interfaces, abstract classes, static factories, concrete adapter drivers, DTOs, and exceptions — infrastructure that doesn't carry business logic. They only each had one real DI-injected service mixed in.
This commit applies the same Services=DI-business-logic / Support=stateless-plumbing rule we've been using throughout the reorg:
**Moved to Support/Integrations/ExchangeRate/** (7 files): ExchangeRateDriver (abstract), ExchangeRateDriverFactory (static), ExchangeRateException, and the four concrete drivers (CurrencyConverter, CurrencyFreak, CurrencyLayer, OpenExchangeRate). These are HTTP adapters over third-party currency APIs — same shape as the Hashids library wrapper classes already in Support.
**Moved to Support/Pdf/** (6 files, merging with existing Pdf utilities): PdfDriver (interface), PdfDriverFactory (static), PdfService (static facade), GotenbergPdfDriver, GotenbergPdfResponse (DTO), ResponseStream (interface). The Support/Pdf/ dir now contains the full PDF rendering subsystem — drivers + sanitizer + template/image utilities.
**Promoted to Services/ root** (the real DI services): ExchangeRateProviderService (CRUD for ExchangeRateProvider model) and FontService (font package install/download orchestration). Both are proper DI services — instance methods, model writes, HTTP side effects.
Services/Integrations/ and Services/Pdf/ are now empty and deleted. Services/ holds only DI-injected classes; Support/ holds all the plumbing.
17 files renamed (git detects 90-99% similarity), 4 consumer files updated (DriverRegistryProvider, PdfServiceProvider, ExchangeRateProviderController, FontController, GeneratesPdfTrait, test). 350 tests pass, Pint clean.
Two follow-ups to the Services reorg that landed in 6d1816bd.
**Documents → Document** (singular). Documents/ was the only plural subdir in app/Services/ — every other bucket (Company, Mail, Module, Pdf, Storage, Update, ExchangeRate) was singular. Renaming to Document/ normalizes the whole tree.
**ExchangeRate → Integrations/ExchangeRate**. Introduces Integrations/ as an umbrella for external-service adapter subsystems. Exchange rate providers move in first; AI providers, payment gateways, and any other driver-pattern integrations land as sibling subdirs (Integrations/Ai/, Integrations/Payment/) without another reorg. Integrations/ was chosen over Providers/ to avoid conceptual collision with Laravel's app/Providers/ — 'check the providers' shouldn't be ambiguous.
17 files moved, 21 consumer files rewritten to point at the new namespaces via literal-string replacement (same approach as the previous reorg). 350 tests pass, Pint clean.
The app/Services/ directory had grown into 22 flat files at the root plus 7 uneven subdirectories — finding anything required scrolling through an alphabetical mix of small CRUD services, infrastructure drivers, and install-time utilities. This commit groups services by domain, folds Backup into a new Storage namespace, and moves framework-infrastructure and install-time helpers out of Services and into Support where they belong.
New Services layout: Documents/ (Invoice, Estimate, RecurringInvoice, Payment, Expense, Transaction, DocumentItem, SerialNumber, Currency — matches the 'Documents' navigation group); Company/ (Company, Member, Invitation); Mail/ (MailConfiguration, CompanyMailConfig); Storage/ (FileDisk, plus Backup folded in). ExchangeRateProviderService moves next to its drivers in ExchangeRate/; FontService moves into Pdf/ where it belongs. CustomerService, ItemService, CustomFieldService stay at the Services root as standalone single-file domains.
Moves to Support/: Hashids/ (library wrapper — not business logic); Setup/ (one-shot install-time utilities — stateless helpers); Pdf/ (ImageUtils, PdfTemplateUtils, plus the existing PdfHtmlSanitizer consolidated into the same subdir). These are all framework infrastructure and stateless utilities — the 'service' label never really fit them.
Namespace declarations in 29 moved files updated to match new paths. 62 consumer files (controllers, other services, tests, database factories, seeders, routes, bootstrap/providers.php) have their use statements rewritten via a literal-string replacement script — no regex meant no risk of half-matching. Three Documents services needed an explicit 'use App\Services\Mail\CompanyMailConfigService' added because the same-namespace short reference they relied on no longer resolves after the split.
Verified: composer dump-autoload, 350 tests pass (850 assertions), vendor/bin/pint clean, npm run build succeeds.
Exchange rate providers are now pluggable via the module Registry. The four built-in drivers (currency_converter, currency_freak, currency_layer, open_exchange_rate) move from a static config array into App\\Providers\\DriverRegistryProvider, which calls Registry::registerExchangeRateDriver() for each during app boot with metadata the frontend needs: label (i18n key), website (help-text URL), and config_fields (schema for driver-specific driver_config JSON).
The Currency Converter's server-type selector and dedicated URL field — previously hardcoded in ExchangeRateProviderModal.vue — are now just another config_fields entry with a visible_when rule that shows the URL input only when type=DEDICATED. Any module that wants to ship a custom driver gets the same treatment for free: declare config_fields in the registration, and the host app's modal renders them automatically.
ExchangeRateDriverFactory::make() falls back to Registry::driverMeta() when a name isn't in the local built-in map, and availableDrivers() merges both sources. ConfigController handles the exchange_rate_drivers key specially by mapping Registry::allDrivers('exchange_rate') to enriched option objects, so the config-file route still works for every other key. The static exchange_rate_drivers + currency_converter_servers arrays in config/invoiceshelf.php are deleted.
Unit tests cover the new Registry::register/flushDrivers, the factory merging built-ins with Registry-contributed drivers, and the factory rejecting unknown names. A feature test exercises the end-to-end /api/v1/config?key=exchange_rate_drivers response shape.
NOTE: this commit depends on invoiceshelf/modules package commit e44d951 which adds the Registry driver API. The package needs to be released and pinned in composer.json before a fresh composer install on this commit will work.
Closes the residual surface from the three published SSRF advisories (GHSA-pc5v-8xwc-v9xq, GHSA-38hf-fq8x-q49r, GHSA-q9wx-ggwq-mcgh / CVE-2026-34365 to 34367) that the original 2.2.0 fix only covered for the Notes field. The same blade templates render company/billing/shipping address fields with {!! !!} via Invoice/Estimate/Payment::getCompanyAddress(), getCustomerBillingAddress(), getCustomerShippingAddress() — and those flow through GeneratesPdfTrait::getFormattedString() which did not call PdfHtmlSanitizer.
Customer-controlled fields (name, street, phone, custom-field values) are substituted into address templates via getFieldsArray() without HTML-escaping, so a malicious customer name like "Acme <img src='http://attacker/probe'>" reaches Dompdf as raw HTML through the address path. Today this is blocked only by the secondary defense of dompdf's enable_remote=false; if a self-hoster sets DOMPDF_ENABLE_REMOTE=true for legitimate remote logos, the address surface immediately re-opens.
Move PdfHtmlSanitizer::sanitize() into the chokepoint at GeneratesPdfTrait::getFormattedString() so all four sinks — notes plus the three address fields, on all three models — get the same treatment via a single call site. v3.0's models (Invoice, Estimate, Payment) already had the simpler getNotes() shape (no per-method PdfHtmlSanitizer wrapper), so the trait edit alone is sufficient — no model edits required on this branch. Verified getFormattedString() is only called from PDF code paths (no email body callers, which use strtr() directly).
This is the v3.0 counterpart to master's f387e751. Re-implemented directly on v3.0 instead of cherry-picked because the import-block divergence from the larger v3.0 refactor produced four merge conflicts that were noisier than just porting the chokepoint change manually.
Extends tests/Unit/PdfHtmlSanitizerTest.php with three new cases covering the address-template scenario, iframe/link tag stripping, and on* event handler removal. All 8 tests pass via vendor/bin/pest tests/Unit/PdfHtmlSanitizerTest.php.
Add IdnEmail validation rule that converts IDN domains to Punycode
via idn_to_ascii() before validating with FILTER_VALIDATE_EMAIL.
Applied to all email fields: customers, members, profiles, admin
users, customer portal profiles, and mail configuration.
Includes unit tests for standard emails, IDN emails, and invalid
inputs.
Fixes#388
Relocate all 14 files from the catch-all app/Space namespace into proper
locations: data providers to Support/Formatters, installation utilities to
Services/Installation, PDF utils to Services/Pdf, module/update classes to
Services/Module and Services/Update, SiteApi trait to Traits, and helpers
to Support.
Extract ~1,400 lines of business logic from 8 fat models (Invoice, Payment,
Estimate, RecurringInvoice, Company, Customer, Expense, User) into 9 new
service classes with constructor injection. Controllers now depend on
services instead of calling static model methods. Shared item/tax creation
logic consolidated into DocumentItemService.
* Convert string references to `::class`
PHP 5.5.9 adds the new static `class` property which provides the fully qualified class name. This is preferred over using strings for class names since the `class` property references are checked by PHP.
* Use Faker methods
Accessing Faker properties was deprecated in Faker 1.14.
* Convert route options to fluent methods
Laravel 8 adopts the tuple syntax for controller actions. Since the old options array is incompatible with this syntax, Shift converted them to use modern, fluent methods.
* Adopt class based routes
* Remove default `app` files
* Shift core files
* Streamline config files
* Set new `ENV` variables
* Default new `bootstrap/app.php`
* Re-register HTTP middleware
* Consolidate service providers
* Re-register service providers
* Re-register routes
* Re-register scheduled commands
* Bump Composer dependencies
* Use `<env>` tags for configuration
`<env>` tags have a lower precedence than system environment variables making it easier to overwrite PHPUnit configuration values in additional environments, such a CI.
Review this blog post for more details on configuration precedence when testing Laravel: https://jasonmccreary.me/articles/laravel-testing-configuration-precedence/
* Adopt anonymous migrations
* Rename `password_resets` table
* Convert `$casts` property to method
* Adopt Laravel type hints
* Mark base controller as `abstract`
* Remove `CreatesApplication` testing trait
* Shift cleanup
* Fix shift first issues
* Updating Rules for laravel 11, sanctum config and pint
* Fix Carbon issue on dashboard
* Temporary fix for tests while migration is issue fixed on laravel side
* Carbon needs numerical values, not strings
* Minimum php version
* Fix domain installation step not fetching the correct company_id
* Fix Role Policy wasn't properly registered
---------
* Create PHP CS Fixer config and add to CI workflow
* Run php cs fixer on project
* Add newline at end of file
* Update to use PHP CS Fixer v3
* Run v3 config on project
* Run seperate config in CI
