feat(benchmark-harness): add Rust fixture generation and size measurement

- Add generate.rs module for fixture generation from vendored test docs
- Add sizes.rs module for framework installation size measurement
- Add CLI commands: generate-fixtures, measure-framework-sizes
- Add 554 benchmark fixture JSON files across 20 file formats
- Update framework_sizes.json with 27 framework configurations
- Fix ground_truth schema to use object format {text_file, source}

Replaces Python scripts with proper Rust implementation including tests.

Author: Goldziher

.github/workflows/benchmarks.yaml

@@ -34,6 +34,8 @@ env:
   RUSTFLAGS: "-C strip=symbols"
   MEASURE_QUALITY: "true"
   OCR_ENABLED: "true"
+  RUN_OCR_BENCHMARKS: "true"
+  GROUND_TRUTH_DIR: "test_documents/ground_truth"
 
 permissions:
   contents: read
@@ -43,8 +45,27 @@ defaults:
     shell: bash
 
 jobs:
+  validate-ground-truth:
+    name: Validate Ground Truth
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          ref: ${{ github.event.inputs.branch || github.ref }}
+
+      - name: Setup Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.11"
+
+      - name: Validate ground truth organization
+        run: python tools/benchmark-harness/scripts/validate_ground_truth.py
+
   setup:
     name: Build harness + native libs
+    needs: validate-ground-truth
     runs-on: ubuntu-latest
     timeout-minutes: 360
     permissions:
@@ -1960,6 +1981,17 @@ jobs:
           echo "Found $ARTIFACT_COUNT artifact directories"
           find benchmark-artifacts -mindepth 1 -maxdepth 1 -type d -exec basename {} \;
 
+      - name: Setup Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.11"
+
+      - name: Measure framework sizes
+        continue-on-error: true
+        run: |
+          python tools/benchmark-harness/scripts/measure_framework_sizes.py \
+            --output consolidated-output/framework-sizes.json || true
+
       - name: Consolidate results
         run: |
           set -euo pipefail

Cargo.lock

@@ -0,0 +1,24 @@
+Sample Document Title
+
+Section 1
+
+This is some introductory text in section 1.
+
+Subsection 1.1
+
+- First list item
+
+- Second list item
+
+This is some introductory text in section 1.1.
+
+- - A dash list item
+
+Section 2
+
+This is some text in section 2.
+
+ Header 1 Header 2 
+ 
+ Value 1 Value 2 
+ Value 3 Value 4

test_documents/ground_truth/asciidoc/test_01.txt

@@ -0,0 +1,8 @@
+{
+  "source": "docling_export",
+  "source_file": "vendored/docling/groundtruth/docling_v2/test_01.asciidoc.md",
+  "document_name": "test_01",
+  "original_md_filename": "test_01.asciidoc",
+  "file_type": "asciidoc",
+  "original_source_doc": null
+}

test_documents/ground_truth/asciidoc/test_01_meta.json

@@ -0,0 +1,78 @@
+2nd Sample Document Title
+
+This is an abstract.
+
+ Section 1: Testing nestedlists
+
+ - First item
+ - Nested item 1
+ - Nested item 2
+ - Second item
+ - Nested ordered item 1
+ - Nested ordered item 2
+ - Deeper nested unordered item
+ - Third item
+ - Nested ordered item 1
+ - Nested ordered item 2
+ - Deeper nested unordered item
+ - Nested ordered item 2
+
+ Section 2
+
+bla bla
+
+bla bla bla
+
+ Section 3: test image
+
+image::images/example1.png[Example Image, width=200, height=150, align=center]
+
+.An example caption for the image
+
+image::images/example2.png[Example Image, width=200, height=150, align=center]
+
+ Section 4: test tables
+
+ Header 1 Header 2 
+ 
+ Value 1 Value 2 
+ Value 3 Value 4 
+
+.Caption for the table 1
+
+ ===
+
+ Header 1 Header 2 
+ 
+ Value 1 Value 2 
+ Value 3 Value 4 
+
+.Caption for the table 2
+
+ ===
+
+ Column 1 Heading Column 2 Heading Column 3 Heading 
+ 
+ Cell 1 Cell 2 Cell 3 
+ Cell 4 Cell 5 colspan=2 Cell spans two columns 
+
+.Caption for the table 3
+
+ ===
+
+ Column 1 Heading Column 2 Heading Column 3 Heading 
+ 
+ Rowspan=2 Cell 2 Cell 3 
+ Cell 5 Cell 6 
+
+.Caption for the table 4
+
+ ===
+
+ Col 1 Col 2 Col 3 Col 4 
+ 
+ Rowspan=2.Colspan=2 Cell spanning 2 rows and 2 columns Col 3 Col 4 
+ Col 3 Col 4 
+ Col 1 Col 2 Col 3 Col 4 
+
+ SubSubSection 2.1.1

test_documents/ground_truth/asciidoc/test_02.txt

@@ -0,0 +1,8 @@
+{
+  "source": "docling_export",
+  "source_file": "vendored/docling/groundtruth/docling_v2/test_02.asciidoc.md",
+  "document_name": "test_02",
+  "original_md_filename": "test_02.asciidoc",
+  "file_type": "asciidoc",
+  "original_source_doc": null
+}

test_documents/ground_truth/asciidoc/test_02_meta.json

@@ -0,0 +1,23 @@
+:\mod-docs-content-type: PROCEDURE :experimental:
+
+Renaming a bookmark
+
+[id="renaming-a-bookmark\{context}"]
+
+You can rename a bookmark to distinguish it from other bookmarks. If you have bookmarks to several folders that all share the same name, you can tell the bookmarks apart if you rename them.
+
+Renaming the bookmark does not rename the folder.
+
+- Check that the side bar lists the bookmark under the new name.
+
+Procedure . Right-click the bookmark in the side bar. . Select Rename…. +
+
+<!-- image -->
+
+ In the Name field, enter the new name for the bookmark. +
+
+<!-- image -->
+
+ Click btn:[Rename]. .Verification
+
+<!-- image -->

test_documents/ground_truth/asciidoc/test_03.txt

@@ -0,0 +1,8 @@
+{
+  "source": "docling_export",
+  "source_file": "vendored/docling/groundtruth/docling_v2/test_03.asciidoc.md",
+  "document_name": "test_03",
+  "original_md_filename": "test_03.asciidoc",
+  "file_type": "asciidoc",
+  "original_source_doc": null
+}

test_documents/ground_truth/asciidoc/test_03_meta.json

@@ -0,0 +1,6 @@
+1 2 3 4 
+ 
+ a b c d 
+ a , c d 
+ a b c d 
+ a b c d

test_documents/ground_truth/csv/csv-comma-in-cell.txt

@@ -0,0 +1,8 @@
+{
+  "source": "docling_export",
+  "source_file": "vendored/docling/groundtruth/docling_v2/csv-comma-in-cell.csv.md",
+  "document_name": "csv-comma-in-cell",
+  "original_md_filename": "csv-comma-in-cell.csv",
+  "file_type": "csv",
+  "original_source_doc": "vendored/docling/csv/csv-comma-in-cell.csv"
+}