trustrender 0.3.3__tar.gz → 0.3.5__tar.gz

{trustrender-0.3.3/src/trustrender.egg-info → trustrender-0.3.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: trustrender
-Version: 0.3.3
+Version: 0.3.5
 Summary: Validate and normalize Stripe, Shopify, and custom billing data for Factur-X/ZUGFeRD
 License-Expression: MIT
 Project-URL: Homepage, https://trustrender.dev
@@ -42,6 +42,11 @@ Dynamic: license-file
 
 # TrustRender
 
+[![PyPI](https://img.shields.io/pypi/v/trustrender.svg)](https://pypi.org/project/trustrender/)
+[![Python](https://img.shields.io/pypi/pyversions/trustrender.svg)](https://pypi.org/project/trustrender/)
+[![Tests](https://img.shields.io/github/actions/workflow/status/verityengine/trustrender/ci.yml?branch=main&label=tests)](https://github.com/verityengine/trustrender/actions)
+[![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](https://opensource.org/licenses/MIT)
+
 Validate and normalize billing data before Factur-X/ZUGFeRD embedding.
 
 ## Quick start
@@ -59,7 +64,7 @@ Invoice:   INV-2026-0187
 From:
 To:        Rheingold Maschinenbau GmbH
 Items:     3
-Total:     $2,685.37
+Total:     $2,945.25
 
 BLOCKED — 1 problem(s)
 
@@ -95,7 +100,7 @@ Invoice:   INV-2026-0187
 From:      NovaTech Solutions GmbH
 To:        Rheingold Maschinenbau GmbH
 Items:     3
-Total:     $2,685.37
+Total:     $2,945.25
 
 PASS — invoice data is valid
 
@@ -149,12 +154,14 @@ Requires Python 3.11+.
 from trustrender import validate_invoice
 from trustrender.adapters import from_stripe
 
-# Raw Stripe API response → validated canonical invoice
-result = validate_invoice(from_stripe(raw_stripe_response), zugferd=True)
+# Raw Stripe API response → canonical invoice (structural validation)
+result = validate_invoice(from_stripe(raw_stripe_response))
 
-if result["render_ready"] and result.get("zugferd_ready"):
+if result["render_ready"]:
     canonical = result["canonical"]
-    # safe to hand off to factur-x / drafthorse
+    # canonical is now ready for either:
+    #   - your own template renderer (most users)
+    #   - the to_zugferd_data() bridge → drafthorse / factur-x (see end-to-end example below)
 else:
     for error in result["errors"]:
         print(f"BLOCKED: {error['message']}")
@@ -167,7 +174,8 @@ else:
 - `errors`: list of blocking issues with rule_id, path, expected/actual
 - `warnings`: advisory issues
 - `normalizations`: field-level provenance (what was renamed, coerced, computed)
-- `zugferd_ready`: bool (if `zugferd=True`)
+
+The `zugferd=True` flag is also accepted but only useful if your input is already in ZUGFeRD shape (with `seller`/`buyer`/`tax_entries`/`payment` keys). For Stripe and Shopify, use the `to_zugferd_data()` bridge after `validate_invoice()` instead — see the end-to-end example below.
 
 Also works with messy data from any source:
 
@@ -213,6 +221,51 @@ result = validate_invoice(from_shopify(raw_shopify_order))
 
 Parses string amounts to floats, combines first_name + last_name, maps order fields to invoice structure, preserves structured address fields. Shopify orders have no seller info or due date — TrustRender flags the missing seller and handles the absent due date correctly.
 
+## End-to-end example: Stripe → real Factur-X PDF
+
+[`examples/with_drafthorse_facturx.py`](examples/with_drafthorse_facturx.py) runs the full pipeline:
+
+```python
+from trustrender import validate_invoice
+from trustrender.adapters import from_stripe
+from trustrender.zugferd import to_zugferd_data, build_invoice_xml, apply_zugferd
+
+# 1. Adapt + validate the Stripe payload (canonical structure check)
+result = validate_invoice(from_stripe(stripe_invoice))
+
+# 2. Bridge canonical → ZUGFeRD shape with the three things Stripe never includes:
+#    seller VAT/address, payment IBAN, applicable tax rate
+zugferd_data = to_zugferd_data(
+    result["canonical"],
+    seller={"name": "...", "address": "...", "city": "...",
+            "postal_code": "...", "country": "DE", "vat_id": "DE..."},
+    payment={"means": "credit_transfer", "iban": "DE..."},
+    tax_rate=19,
+)
+# Real EN 16931 contract validation runs here — catches anything XML build would reject.
+
+# 3. drafthorse builds UN/CEFACT CII XML
+xml_bytes = build_invoice_xml(zugferd_data)
+
+# 4. factur-x embeds the XML into a PDF/A-3b container
+factur_x_pdf = apply_zugferd(your_visual_pdf_bytes, xml_bytes, lang="de")
+```
+
+Output: a 15 KB Factur-X invoice PDF that validates against the EN 16931 XSD ([sample committed at `examples/invoice_facturx.pdf`](examples/invoice_facturx.pdf)).
+
+```
+$ python examples/with_drafthorse_facturx.py
+Step 1: Adapt + validate via TrustRender (canonical structure)  → status=ready
+Step 2: Bridge canonical → ZUGFeRD shape (one call)             → passes EN 16931 contract validation
+Step 3: Build CII XML via drafthorse                            → 8,312 bytes of CII XML
+Step 4: Render visual PDF (your template)                       → 1,599 bytes
+Step 5: Embed CII XML as PDF/A-3b                               → 15,004 bytes
+Step 6: Verify with factur-x library                            → ✓ passes EN 16931 XSD
+✓ factur-x.xml is embedded in the PDF
+```
+
+What's TrustRender vs what's you: the seller VAT, payment IBAN, and tax rate come from your billing setup — Stripe's API doesn't include them, and TrustRender doesn't invent them. Everything else is automatic.
+
 ## What it normalizes
 
 90+ vendor field aliases across QuickBooks, Xero, Stripe, and generic CSV/ERP formats:
@@ -257,53 +310,47 @@ trustrender serve --templates <dir> [--port 8190]
 trustrender doctor [--smoke]
 ```
 
-## Integration with factur-x / drafthorse
-
-TrustRender validates and normalizes. You generate and embed with the library of your choice.
-
-```python
-from trustrender import validate_invoice
-
-# Step 1: Validate with TrustRender
-result = validate_invoice(messy_data, zugferd=True)
-if not result["render_ready"] or not result["zugferd_ready"]:
-    raise ValueError(f"Invoice blocked: {result['errors']}")
-
-# Step 2: Use the canonical payload with drafthorse or factur-x
-canonical = result["canonical"]
-# ... your existing ZUGFeRD generation code here
-```
-
 ## EN 16931 e-invoicing (narrow scope)
 
-Catches many document-level and ZUGFeRD/EN 16931 readiness issues before embedding. Currently supports:
+Catches structural and contract-level EN 16931 issues before drafthorse builds the XML. Currently supports:
 
 - **Domestic German B2B invoices** with standard VAT, EUR, SEPA payment
-- Single or mixed VAT rates (7% + 19%)
+- Single VAT rate per invoice (mixed rates not yet supported in the bridge)
 - Invoice type 380 and credit note 381
-- PDF/A-3b with embedded CII XML (requires `trustrender[render]`)
+- PDF/A-3b with embedded CII XML via `apply_zugferd()` (wraps `factur-x.attach_xml`)
+
+Not supported, fails loudly: reverse charge, cross-border, allowances/charges, non-EUR currencies, non-DE seller country.
 
-Not supported (fails loudly): reverse charge, cross-border, allowances/charges, non-EUR currencies.
+What still slips past TrustRender: things only the official Schematron rules catch (cross-field business rules, value range checks). Run [`xml_check_xsd`](https://github.com/akretion/factur-x) and `xml_check_schematron` on the produced XML for full compliance certainty.
 
 See [docs/einvoice-scope.md](docs/einvoice-scope.md) for the full scope matrix.
 
-## Optional: PDF rendering
+## Also: deterministic PDF rendering
+
+TrustRender ships a Jinja2 + Typst render engine alongside the validation layer. Used internally to produce the visual PDF for Factur-X embedding, and usable standalone if you want one Python package that goes from data → validated canonical → final PDF without bringing in a headless browser.
 
-If you also want TrustRender to generate PDFs (not just validate):
+- 1,000-line invoice rendered in ~59 ms median (~72 ms with content varying every render; Apple M5 Pro, single core, v0.3.5, 2026-08-29)
+- 1,011-page invoice rendered in 37 seconds
+- No headless browser, no Chromium binary, no Java runtime — pure Python + the Typst CLI
+- SHA-256 fingerprinting of input, template, and output for audit trails (see `trustrender audit`)
+- Drift detection vs saved baselines for visual regression in CI
+- Subprocess-based backend with real timeouts (killable renders, not just abandoned)
+- Bundled fonts, six built-in templates (invoice, einvoice, statement, letter, receipt, report)
 
 ```
 pip install "trustrender[render]"
 trustrender render invoice.j2.typ data.json -o invoice.pdf --zugferd en16931
 ```
 
-Rendering uses Typst — no browser, no Chromium. Fast and deterministic.
+`render`, `preflight`, `audit`, `baseline`, and `serve` are all part of the rendering surface — see `trustrender --help` for the full set.
 
 ## What this is not
 
 - Not a full AP automation platform
-- Not an e-invoice compliance certification
-- Not an AI-powered data fixer (all corrections are deterministic)
-- Not a replacement for factur-x or drafthorse — it's the validation layer you run before them
+- Not an e-invoice compliance certification (run Schematron on the output XML for that)
+- Not an AI-powered data fixer — all normalizations and validations are deterministic
+- Not a replacement for factur-x or drafthorse — it validates billing data before them and wraps their entry points for convenience
+- Not a billing platform replacement — it expects you already collect payment via Stripe, Shopify, or something else
 
 ## Development
 

{trustrender-0.3.3 → trustrender-0.3.5}/README.md

@@ -1,5 +1,10 @@
 # TrustRender
 
+[![PyPI](https://img.shields.io/pypi/v/trustrender.svg)](https://pypi.org/project/trustrender/)
+[![Python](https://img.shields.io/pypi/pyversions/trustrender.svg)](https://pypi.org/project/trustrender/)
+[![Tests](https://img.shields.io/github/actions/workflow/status/verityengine/trustrender/ci.yml?branch=main&label=tests)](https://github.com/verityengine/trustrender/actions)
+[![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](https://opensource.org/licenses/MIT)
+
 Validate and normalize billing data before Factur-X/ZUGFeRD embedding.
 
 ## Quick start
@@ -17,7 +22,7 @@ Invoice:   INV-2026-0187
 From:
 To:        Rheingold Maschinenbau GmbH
 Items:     3
-Total:     $2,685.37
+Total:     $2,945.25
 
 BLOCKED — 1 problem(s)
 
@@ -53,7 +58,7 @@ Invoice:   INV-2026-0187
 From:      NovaTech Solutions GmbH
 To:        Rheingold Maschinenbau GmbH
 Items:     3
-Total:     $2,685.37
+Total:     $2,945.25
 
 PASS — invoice data is valid
 
@@ -107,12 +112,14 @@ Requires Python 3.11+.
 from trustrender import validate_invoice
 from trustrender.adapters import from_stripe
 
-# Raw Stripe API response → validated canonical invoice
-result = validate_invoice(from_stripe(raw_stripe_response), zugferd=True)
+# Raw Stripe API response → canonical invoice (structural validation)
+result = validate_invoice(from_stripe(raw_stripe_response))
 
-if result["render_ready"] and result.get("zugferd_ready"):
+if result["render_ready"]:
     canonical = result["canonical"]
-    # safe to hand off to factur-x / drafthorse
+    # canonical is now ready for either:
+    #   - your own template renderer (most users)
+    #   - the to_zugferd_data() bridge → drafthorse / factur-x (see end-to-end example below)
 else:
     for error in result["errors"]:
         print(f"BLOCKED: {error['message']}")
@@ -125,7 +132,8 @@ else:
 - `errors`: list of blocking issues with rule_id, path, expected/actual
 - `warnings`: advisory issues
 - `normalizations`: field-level provenance (what was renamed, coerced, computed)
-- `zugferd_ready`: bool (if `zugferd=True`)
+
+The `zugferd=True` flag is also accepted but only useful if your input is already in ZUGFeRD shape (with `seller`/`buyer`/`tax_entries`/`payment` keys). For Stripe and Shopify, use the `to_zugferd_data()` bridge after `validate_invoice()` instead — see the end-to-end example below.
 
 Also works with messy data from any source:
 
@@ -171,6 +179,51 @@ result = validate_invoice(from_shopify(raw_shopify_order))
 
 Parses string amounts to floats, combines first_name + last_name, maps order fields to invoice structure, preserves structured address fields. Shopify orders have no seller info or due date — TrustRender flags the missing seller and handles the absent due date correctly.
 
+## End-to-end example: Stripe → real Factur-X PDF
+
+[`examples/with_drafthorse_facturx.py`](examples/with_drafthorse_facturx.py) runs the full pipeline:
+
+```python
+from trustrender import validate_invoice
+from trustrender.adapters import from_stripe
+from trustrender.zugferd import to_zugferd_data, build_invoice_xml, apply_zugferd
+
+# 1. Adapt + validate the Stripe payload (canonical structure check)
+result = validate_invoice(from_stripe(stripe_invoice))
+
+# 2. Bridge canonical → ZUGFeRD shape with the three things Stripe never includes:
+#    seller VAT/address, payment IBAN, applicable tax rate
+zugferd_data = to_zugferd_data(
+    result["canonical"],
+    seller={"name": "...", "address": "...", "city": "...",
+            "postal_code": "...", "country": "DE", "vat_id": "DE..."},
+    payment={"means": "credit_transfer", "iban": "DE..."},
+    tax_rate=19,
+)
+# Real EN 16931 contract validation runs here — catches anything XML build would reject.
+
+# 3. drafthorse builds UN/CEFACT CII XML
+xml_bytes = build_invoice_xml(zugferd_data)
+
+# 4. factur-x embeds the XML into a PDF/A-3b container
+factur_x_pdf = apply_zugferd(your_visual_pdf_bytes, xml_bytes, lang="de")
+```
+
+Output: a 15 KB Factur-X invoice PDF that validates against the EN 16931 XSD ([sample committed at `examples/invoice_facturx.pdf`](examples/invoice_facturx.pdf)).
+
+```
+$ python examples/with_drafthorse_facturx.py
+Step 1: Adapt + validate via TrustRender (canonical structure)  → status=ready
+Step 2: Bridge canonical → ZUGFeRD shape (one call)             → passes EN 16931 contract validation
+Step 3: Build CII XML via drafthorse                            → 8,312 bytes of CII XML
+Step 4: Render visual PDF (your template)                       → 1,599 bytes
+Step 5: Embed CII XML as PDF/A-3b                               → 15,004 bytes
+Step 6: Verify with factur-x library                            → ✓ passes EN 16931 XSD
+✓ factur-x.xml is embedded in the PDF
+```
+
+What's TrustRender vs what's you: the seller VAT, payment IBAN, and tax rate come from your billing setup — Stripe's API doesn't include them, and TrustRender doesn't invent them. Everything else is automatic.
+
 ## What it normalizes
 
 90+ vendor field aliases across QuickBooks, Xero, Stripe, and generic CSV/ERP formats:
@@ -215,53 +268,47 @@ trustrender serve --templates <dir> [--port 8190]
 trustrender doctor [--smoke]
 ```
 
-## Integration with factur-x / drafthorse
-
-TrustRender validates and normalizes. You generate and embed with the library of your choice.
-
-```python
-from trustrender import validate_invoice
-
-# Step 1: Validate with TrustRender
-result = validate_invoice(messy_data, zugferd=True)
-if not result["render_ready"] or not result["zugferd_ready"]:
-    raise ValueError(f"Invoice blocked: {result['errors']}")
-
-# Step 2: Use the canonical payload with drafthorse or factur-x
-canonical = result["canonical"]
-# ... your existing ZUGFeRD generation code here
-```
-
 ## EN 16931 e-invoicing (narrow scope)
 
-Catches many document-level and ZUGFeRD/EN 16931 readiness issues before embedding. Currently supports:
+Catches structural and contract-level EN 16931 issues before drafthorse builds the XML. Currently supports:
 
 - **Domestic German B2B invoices** with standard VAT, EUR, SEPA payment
-- Single or mixed VAT rates (7% + 19%)
+- Single VAT rate per invoice (mixed rates not yet supported in the bridge)
 - Invoice type 380 and credit note 381
-- PDF/A-3b with embedded CII XML (requires `trustrender[render]`)
+- PDF/A-3b with embedded CII XML via `apply_zugferd()` (wraps `factur-x.attach_xml`)
+
+Not supported, fails loudly: reverse charge, cross-border, allowances/charges, non-EUR currencies, non-DE seller country.
 
-Not supported (fails loudly): reverse charge, cross-border, allowances/charges, non-EUR currencies.
+What still slips past TrustRender: things only the official Schematron rules catch (cross-field business rules, value range checks). Run [`xml_check_xsd`](https://github.com/akretion/factur-x) and `xml_check_schematron` on the produced XML for full compliance certainty.
 
 See [docs/einvoice-scope.md](docs/einvoice-scope.md) for the full scope matrix.
 
-## Optional: PDF rendering
+## Also: deterministic PDF rendering
+
+TrustRender ships a Jinja2 + Typst render engine alongside the validation layer. Used internally to produce the visual PDF for Factur-X embedding, and usable standalone if you want one Python package that goes from data → validated canonical → final PDF without bringing in a headless browser.
 
-If you also want TrustRender to generate PDFs (not just validate):
+- 1,000-line invoice rendered in ~59 ms median (~72 ms with content varying every render; Apple M5 Pro, single core, v0.3.5, 2026-08-29)
+- 1,011-page invoice rendered in 37 seconds
+- No headless browser, no Chromium binary, no Java runtime — pure Python + the Typst CLI
+- SHA-256 fingerprinting of input, template, and output for audit trails (see `trustrender audit`)
+- Drift detection vs saved baselines for visual regression in CI
+- Subprocess-based backend with real timeouts (killable renders, not just abandoned)
+- Bundled fonts, six built-in templates (invoice, einvoice, statement, letter, receipt, report)
 
 ```
 pip install "trustrender[render]"
 trustrender render invoice.j2.typ data.json -o invoice.pdf --zugferd en16931
 ```
 
-Rendering uses Typst — no browser, no Chromium. Fast and deterministic.
+`render`, `preflight`, `audit`, `baseline`, and `serve` are all part of the rendering surface — see `trustrender --help` for the full set.
 
 ## What this is not
 
 - Not a full AP automation platform
-- Not an e-invoice compliance certification
-- Not an AI-powered data fixer (all corrections are deterministic)
-- Not a replacement for factur-x or drafthorse — it's the validation layer you run before them
+- Not an e-invoice compliance certification (run Schematron on the output XML for that)
+- Not an AI-powered data fixer — all normalizations and validations are deterministic
+- Not a replacement for factur-x or drafthorse — it validates billing data before them and wraps their entry points for convenience
+- Not a billing platform replacement — it expects you already collect payment via Stripe, Shopify, or something else
 
 ## Development
 

{trustrender-0.3.3 → trustrender-0.3.5}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "trustrender"
-version = "0.3.3"
+version = "0.3.5"
 description = "Validate and normalize Stripe, Shopify, and custom billing data for Factur-X/ZUGFeRD"
 license = "MIT"
 classifiers = [
@@ -66,8 +66,14 @@ where = ["src"]
 trustrender = ["playground/**/*", "fonts/**/*", "builtin_templates/*"]
 
 [tool.ruff]
-line-length = 99
+line-length = 140
 target-version = "py311"
 
 [tool.ruff.lint]
 select = ["E", "F", "I", "W"]
+
+[tool.ruff.lint.per-file-ignores]
+# Dashboard generates HTML/CSS as inline string literals — long lines are intentional
+"src/trustrender/dashboard.py" = ["E501"]
+# Tests: long fixture strings, side-effect render() calls, ad-hoc fixture vars
+"tests/*" = ["E501", "F841"]

{trustrender-0.3.3 → trustrender-0.3.5}/src/trustrender/__init__.py

@@ -23,6 +23,7 @@ from .invoice_ingest import IngestionReport, ingest_invoice
 try:
     from .engine import CompileBackend, compile_typst, compile_typst_file
     from .templates import render_template
+
     _HAS_RENDER = True
 except ImportError:
     _HAS_RENDER = False
@@ -36,22 +37,29 @@ class RenderResult:
     trace_id: str | None = None
 
 
-__version__ = "0.3.3"
+__version__ = "0.3.5"
 
 __all__ = [
     # Core API (always available)
-    "validate_invoice", "ingest_invoice", "IngestionReport",
+    "validate_invoice",
+    "ingest_invoice",
+    "IngestionReport",
     # Rendering API (requires [render] extras)
-    "render", "audit", "AuditResult",
+    "render",
+    "audit",
+    "AuditResult",
     # Errors
-    "TrustRenderError", "ErrorCode",
+    "TrustRenderError",
+    "ErrorCode",
     # Meta
-    "__version__", "bundled_font_dir",
+    "__version__",
+    "bundled_font_dir",
 ]
 
 
 # ── Core API: validation (always available) ──────────────────────────
 
+
 def validate_invoice(data: dict, *, zugferd: bool = False) -> dict:
     """Validate and normalize messy invoice data. Returns structured result.
 
@@ -100,10 +108,9 @@ def validate_invoice(data: dict, *, zugferd: bool = False) -> dict:
 
     if zugferd and report.render_ready:
         from .zugferd import validate_zugferd_invoice_data
+
         zug_errors = validate_zugferd_invoice_data(report.canonical)
-        result["zugferd_errors"] = [
-            {"path": e.path, "message": e.message} for e in zug_errors
-        ]
+        result["zugferd_errors"] = [{"path": e.path, "message": e.message} for e in zug_errors]
         result["zugferd_ready"] = len(zug_errors) == 0
     elif zugferd:
         result["zugferd_errors"] = []
@@ -171,8 +178,7 @@ def _require_render():
     """Raise a clear error if rendering dependencies are not installed."""
     if not _HAS_RENDER:
         raise TrustRenderError(
-            "Rendering requires additional dependencies. "
-            "Install with: pip install trustrender[render]",
+            "Rendering requires additional dependencies. Install with: pip install trustrender[render]",
             code=ErrorCode.BACKEND_ERROR,
             stage="import",
         )
@@ -221,7 +227,7 @@ def _render_document_pipeline(
     trace = RenderTrace(
         template_name=display_name or template_path.name,
         template_hash=f"sha256:{hashlib.sha256(template_path.read_bytes()).hexdigest()[:16]}",
-        data_hash=f"sha256:{hashlib.sha256(json.dumps(data, sort_keys=True, separators=(',',':')).encode()).hexdigest()[:16]}",
+        data_hash=f"sha256:{hashlib.sha256(json.dumps(data, sort_keys=True, separators=(',', ':')).encode()).hexdigest()[:16]}",
         engine_version=__version__,
         zugferd_profile=zugferd or "",
         validated=validate,
@@ -330,16 +336,18 @@ def _render_document_pipeline(
                 timeout=timeout,
                 pdf_standards=pdf_standards,
             )
-        trace.stages.append(StageTrace(
-            stage="compilation",
-            status="pass",
-            duration_ms=int((time.monotonic() - t0) * 1000),
-            metadata={
-                "template_type": "jinja2" if is_jinja else "raw",
-                "pdf_standards": pdf_standards or [],
-                "pdf_size": len(pdf_bytes),
-            },
-        ))
+        trace.stages.append(
+            StageTrace(
+                stage="compilation",
+                status="pass",
+                duration_ms=int((time.monotonic() - t0) * 1000),
+                metadata={
+                    "template_type": "jinja2" if is_jinja else "raw",
+                    "pdf_standards": pdf_standards or [],
+                    "pdf_size": len(pdf_bytes),
+                },
+            )
+        )
 
         # 4. ZUGFeRD post-processing
         if zugferd:
@@ -363,12 +371,14 @@ def _render_document_pipeline(
                     )
 
                 pdf_bytes = apply_zugferd(pdf_bytes, xml_bytes)
-                trace.stages.append(StageTrace(
-                    stage="zugferd_postprocess",
-                    status="pass",
-                    duration_ms=int((time.monotonic() - t0) * 1000),
-                    metadata={"xml_size": len(xml_bytes), "profile": zugferd},
-                ))
+                trace.stages.append(
+                    StageTrace(
+                        stage="zugferd_postprocess",
+                        status="pass",
+                        duration_ms=int((time.monotonic() - t0) * 1000),
+                        metadata={"xml_size": len(xml_bytes), "profile": zugferd},
+                    )
+                )
             except TrustRenderError:
                 raise
             except Exception as exc:
@@ -387,12 +397,14 @@ def _render_document_pipeline(
             t0 = time.monotonic()
             prov_record = create_provenance(template_path, data)
             pdf_bytes = embed_provenance(pdf_bytes, prov_record)
-            trace.stages.append(StageTrace(
-                stage="provenance",
-                status="pass",
-                duration_ms=int((time.monotonic() - t0) * 1000),
-                metadata={"proof_hash": prov_record.proof[:30]},
-            ))
+            trace.stages.append(
+                StageTrace(
+                    stage="provenance",
+                    status="pass",
+                    duration_ms=int((time.monotonic() - t0) * 1000),
+                    metadata={"proof_hash": prov_record.proof[:30]},
+                )
+            )
             trace.provenance_hash = prov_record.proof
 
         # Output fingerprint: hash the final PDF bytes after all post-processing
@@ -545,7 +557,8 @@ def audit(
     import time
 
     from .fingerprint import InputFingerprint, compare, compute_fingerprint
-    from .regression import check_drift, save_baseline as _save_baseline
+    from .regression import check_drift
+    from .regression import save_baseline as _save_baseline
     from .semantic import SemanticReport, validate_semantics
 
     _SUPPORTED_ZUGFERD = {"en16931"}
@@ -623,9 +636,7 @@ def audit(
                 pdf_bytes,
                 zugferd_valid=zugferd is not None,
                 contract_valid=validate,
-                semantic_issue_count=(
-                    len(semantic_report.issues) if semantic_report else 0
-                ),
+                semantic_issue_count=(len(semantic_report.issues) if semantic_report else 0),
             )
 
     # 6. Save baseline (if requested)
@@ -638,9 +649,7 @@ def audit(
             render_duration_ms=render_duration_ms,
             zugferd_valid=zugferd is not None if zugferd else None,
             contract_valid=True if validate else None,
-            semantic_issue_count=(
-                len(semantic_report.issues) if semantic_report else 0
-            ),
+            semantic_issue_count=(len(semantic_report.issues) if semantic_report else 0),
         )
 
     return AuditResult(
@@ -659,8 +668,7 @@ def _resolve_data(data: dict | str | os.PathLike) -> dict:
 
     if not isinstance(data, (str, os.PathLike)):
         raise TrustRenderError(
-            f"Data must be a dict, JSON string, or path to a .json file, "
-            f"got {type(data).__name__}",
+            f"Data must be a dict, JSON string, or path to a .json file, got {type(data).__name__}",
             code=ErrorCode.INVALID_DATA,
             stage="data_resolution",
         )