Nomad changes

fixtures/smoke/docx_basic.json

@@ -0,0 +1,32 @@
+{
+  "id": "smoke_docx_basic",
+  "category": "smoke",
+  "description": "Smoke test: DOCX with formatted text",
+  "tags": ["smoke", "office", "docx"],
+  "input": {
+    "path": "docx/fake.docx",
+    "mime_type": "application/vnd.openxmlformats-officedocument.wordprocessingml.document",
+    "config": {}
+  },
+  "assertions": [
+    {
+      "type": "equals",
+      "field": "mime_type",
+      "value": "application/vnd.openxmlformats-officedocument.wordprocessingml.document"
+    },
+    {
+      "type": "min_length",
+      "field": "content",
+      "value": 20
+    },
+    {
+      "type": "contains_any",
+      "field": "content",
+      "values": ["Lorem", "ipsum", "document", "text"]
+    }
+  ],
+  "skip": {
+    "languages": ["wasm"],
+    "reason": "WASM cannot access filesystem; use extractBytes with file content instead"
+  }
+}

fixtures/smoke/html_basic.json

@@ -0,0 +1,32 @@
+{
+  "id": "smoke_html_basic",
+  "category": "smoke",
+  "description": "Smoke test: HTML table extraction",
+  "tags": ["smoke", "html"],
+  "input": {
+    "path": "html/simple_table.html",
+    "mime_type": "text/html",
+    "config": {}
+  },
+  "assertions": [
+    {
+      "type": "equals",
+      "field": "mime_type",
+      "value": "text/html"
+    },
+    {
+      "type": "min_length",
+      "field": "content",
+      "value": 10
+    },
+    {
+      "type": "contains_any",
+      "field": "content",
+      "values": ["Sample Data Table", "Laptop", "Electronics", "Product"]
+    }
+  ],
+  "skip": {
+    "languages": ["wasm"],
+    "reason": "WASM cannot access filesystem; use extractBytes with file content instead"
+  }
+}

fixtures/smoke/image_png.json

@@ -0,0 +1,24 @@
+{
+  "id": "smoke_image_png",
+  "category": "smoke",
+  "description": "Smoke test: PNG image (without OCR, metadata only)",
+  "tags": ["smoke", "image", "png"],
+  "input": {
+    "path": "images/sample.png",
+    "media_type": "image/png",
+    "config": {
+      "disable_ocr": true
+    }
+  },
+  "assertions": [
+    {
+      "type": "equals",
+      "field": "mime_type",
+      "value": "image/png"
+    }
+  ],
+  "skip": {
+    "languages": ["wasm"],
+    "reason": "WASM cannot access filesystem; use extractBytes with file content instead"
+  }
+}

fixtures/smoke/json_basic.json

@@ -0,0 +1,27 @@
+{
+  "id": "smoke_json_basic",
+  "category": "smoke",
+  "description": "Smoke test: JSON file extraction",
+  "tags": ["smoke", "structured", "json"],
+  "input": {
+    "path": "json/simple.json",
+    "mime_type": "application/json",
+    "config": {}
+  },
+  "assertions": [
+    {
+      "type": "equals",
+      "field": "mime_type",
+      "value": "application/json"
+    },
+    {
+      "type": "min_length",
+      "field": "content",
+      "value": 5
+    }
+  ],
+  "skip": {
+    "languages": ["wasm"],
+    "reason": "WASM cannot access filesystem; use extractBytes with file content instead"
+  }
+}

fixtures/smoke/ocr_image_png.json

@@ -0,0 +1,29 @@
+{
+  "id": "ocr_image_png",
+  "category": "smoke",
+  "description": "OCR: PNG image extraction with OCR enabled. In WASM this exercises the Uint8Array bridge parameter and Promise await in the generated OcrBackend bridge.",
+  "tags": ["smoke", "ocr", "image", "png"],
+  "call": "extract_bytes",
+  "input": {
+    "data": "images/test_hello_world.png",
+    "mime_type": "image/png",
+    "config": {}
+  },
+  "assertions": [
+    {
+      "type": "equals",
+      "field": "mime_type",
+      "value": "image/png"
+    },
+    {
+      "type": "min_length",
+      "field": "content",
+      "value": 1
+    },
+    {
+      "type": "contains_any",
+      "field": "content",
+      "values": ["Hello", "World", "hello", "world"]
+    }
+  ]
+}

fixtures/smoke/pdf_basic.json

@@ -0,0 +1,32 @@
+{
+  "id": "smoke_pdf_basic",
+  "category": "smoke",
+  "description": "Smoke test: PDF with simple text extraction",
+  "tags": ["smoke", "pdf", "basic"],
+  "input": {
+    "path": "pdf/fake_memo.pdf",
+    "mime_type": "application/pdf",
+    "config": {}
+  },
+  "assertions": [
+    {
+      "type": "equals",
+      "field": "mime_type",
+      "value": "application/pdf"
+    },
+    {
+      "type": "min_length",
+      "field": "content",
+      "value": 50
+    },
+    {
+      "type": "contains_any",
+      "field": "content",
+      "values": ["May 5, 2023", "To Whom it May Concern"]
+    }
+  ],
+  "skip": {
+    "languages": ["wasm"],
+    "reason": "WASM cannot access filesystem; use extractBytes with file content instead"
+  }
+}

fixtures/smoke/txt_basic.json

@@ -0,0 +1,27 @@
+{
+  "id": "smoke_txt_basic",
+  "category": "smoke",
+  "description": "Smoke test: Plain text file",
+  "tags": ["smoke", "text", "plaintext"],
+  "input": {
+    "path": "text/report.txt",
+    "mime_type": "text/plain",
+    "config": {}
+  },
+  "assertions": [
+    {
+      "type": "equals",
+      "field": "mime_type",
+      "value": "text/plain"
+    },
+    {
+      "type": "min_length",
+      "field": "content",
+      "value": 5
+    }
+  ],
+  "skip": {
+    "languages": ["wasm"],
+    "reason": "WASM cannot access filesystem; use extractBytes with file content instead"
+  }
+}

fixtures/smoke/xlsx_basic.json

@@ -0,0 +1,57 @@
+{
+  "id": "smoke_xlsx_basic",
+  "category": "smoke",
+  "description": "Smoke test: XLSX with basic spreadsheet data including tables",
+  "tags": ["smoke", "office", "xlsx", "tables"],
+  "input": {
+    "path": "xlsx/stanley_cups.xlsx",
+    "mime_type": "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
+    "config": {}
+  },
+  "assertions": [
+    {
+      "type": "equals",
+      "field": "mime_type",
+      "value": "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet"
+    },
+    {
+      "type": "min_length",
+      "field": "content",
+      "value": 100
+    },
+    {
+      "type": "contains_all",
+      "field": "content",
+      "values": [
+        "Team",
+        "Location",
+        "Stanley Cups",
+        "Blues",
+        "Flyers",
+        "Maple Leafs",
+        "STL",
+        "PHI",
+        "TOR"
+      ]
+    },
+    {
+      "type": "count_min",
+      "field": "tables",
+      "value": 1
+    },
+    {
+      "type": "greater_than_or_equal",
+      "field": "metadata.format.excel.sheet_count",
+      "value": 2
+    },
+    {
+      "type": "contains_all",
+      "field": "metadata.format.excel.sheet_names",
+      "values": ["Stanley Cups"]
+    }
+  ],
+  "skip": {
+    "languages": ["wasm"],
+    "reason": "WASM cannot access filesystem; use extractBytes with file content instead"
+  }
+}