Merge pull request #172 from AdaWorldAPI/claude/ocr-pdf-plan-mints

ogar-vocab: mint the 0x08 OCR PDF-to-text plan concepts (0x0805..0x0809)

Author: AdaWorldAPI

crates/ogar-class-view/src/lib.rs

@@ -77,6 +77,7 @@ use ogar_vocab::{
     automation_trigger,
     billable_work_entry,
     billing_party,
+    blob,
     bone,
     canonical_concept_id,
     charset,
@@ -98,6 +99,7 @@ use ogar_vocab::{
     mars_software,
     medication,
     network_layer,
+    ocr_renderer,
     osm_changeset,
     osm_element_tag,
     osm_gpx_trace,
@@ -108,6 +110,8 @@ use ogar_vocab::{
     osm_user,
     osm_way,
     osm_way_node,
+    page_image,
+    page_layout,
     patient,
     payment_record,
     pricelist,
@@ -141,6 +145,7 @@ use ogar_vocab::{
     recoder,
     skeleton,
     tax_policy,
+    textline,
     treatment,
     unicharset,
     unit_of_measure,
@@ -200,6 +205,11 @@ fn all_canonical_classes() -> Vec<(&'static str, Class)> {
         ("recoder", recoder()),
         ("charset", charset()),
         ("network_layer", network_layer()),
+        ("textline", textline()),
+        ("blob", blob()),
+        ("page_layout", page_layout()),
+        ("page_image", page_image()),
+        ("ocr_renderer", ocr_renderer()),
         // ── 0x09XX — health (OGIT Healthcare) ──
         ("patient", patient()),
         ("diagnosis", diagnosis()),

crates/ogar-vocab/src/lib.rs

@@ -1203,6 +1203,14 @@ const CODEBOOK: &[(&str, u16)] = &[
     ("recoder", 0x0802),
     ("charset", 0x0803),
     ("network_layer", 0x0804),
+    // PDF→text plan mints (tesseract-rs `pdf-to-text-ocr-v1.md` Phase 0 D0.3;
+    // the plan IS the emit-seam declaration — each row names its consumer
+    // phase, honoring the mint-on-emit guard rather than pre-minting blind).
+    ("textline", 0x0805),
+    ("blob", 0x0806),
+    ("page_layout", 0x0807),
+    ("page_image", 0x0808),
+    ("ocr_renderer", 0x0809),
     // ── 0x09XX — Health domain (clinical / patient / care) ──
     // medcare-rs Healthcare-namespace promotion (Northstar T9). The 7
     // entities the OGIT `NTO/Healthcare/entities/` TTL ships, projected
@@ -1678,6 +1686,25 @@ pub mod class_ids {
     /// ruff→OGAR network harvest lands here). Wire-declared first in the
     /// lance-graph contract mirror; this is the authoritative OGAR counterpart.
     pub const NETWORK_LAYER: u16 = 0x0804;
+    /// `textline` (`0x0805`) — a segmented text line: box + baseline, the unit
+    /// the LSTM line recognizer consumes and the word/box extract references
+    /// (plan Phases 1B + 3E; the line images `RecognizeLine` reads).
+    pub const TEXTLINE: u16 = 0x0805;
+    /// `blob` (`0x0806`) — a connected-component blob container (Tesseract
+    /// `C_BLOB`/`BLOBNBOX` kind), the page-segmentation substrate (plan
+    /// Phase 3B/3D). Content (outlines/pixels) stays in content stores.
+    pub const BLOB: u16 = 0x0806;
+    /// `page_layout` (`0x0807`) — the layout-analysis result container:
+    /// blocks / columns / reading order over one page (plan Phase 3F/4A).
+    pub const PAGE_LAYOUT: u16 = 0x0807;
+    /// `page_image` (`0x0808`) — a rasterized page container (decoded +
+    /// grey/binary normalized), the OCR input kind (plan Phase 2).
+    pub const PAGE_IMAGE: u16 = 0x0808;
+    /// `ocr_renderer` (`0x0809`) — an output-rendering KIND (txt / tsv /
+    /// hOCR / searchable-PDF), one slot; the concrete format is the
+    /// classid custom-low half, mirroring the `network_layer` pattern
+    /// (plan Phase 4B-D).
+    pub const OCR_RENDERER: u16 = 0x0809;
 
     // ── 0x09XX — health domain (medcare-rs Healthcare namespace) ──
 
@@ -1878,6 +1905,11 @@ pub mod class_ids {
         ("recoder", RECODER),
         ("charset", CHARSET),
         ("network_layer", NETWORK_LAYER),
+        ("textline", TEXTLINE),
+        ("blob", BLOB),
+        ("page_layout", PAGE_LAYOUT),
+        ("page_image", PAGE_IMAGE),
+        ("ocr_renderer", OCR_RENDERER),
         // 0x09XX — health
         ("patient", PATIENT),
         ("diagnosis", DIAGNOSIS),
@@ -1986,7 +2018,7 @@ pub mod class_ids {
             // lance-graph mirror is rebuilt against it.
             assert_eq!(
                 ALL.len(),
-                79,
+                84,
                 "class_ids::ALL count changed — update this pin AND the \
                  lance-graph mirror COUNT_FUSE (crates/lance-graph-ogar/src/lib.rs) \
                  in the same PR",
@@ -2808,11 +2840,16 @@ pub fn all_promoted_classes() -> Vec<Class> {
         // `osint_system` / `osint_person` mints); no calls follow — OGAR
         // vocabulary carries no OSINT concept names, see the CODEBOOK
         // 0x07XX section note.
-        // 0x08XX — OCR arm (4 container kinds), in class_ids::ALL order.
+        // 0x08XX — OCR arm (9 container kinds), in class_ids::ALL order.
         unicharset(),
         recoder(),
         charset(),
         network_layer(),
+        textline(),
+        blob(),
+        page_layout(),
+        page_image(),
+        ocr_renderer(),
         // 0x09XX — health arm (7 OGIT Healthcare concepts), in
         // class_ids::ALL order.
         patient(),
@@ -3828,6 +3865,69 @@ pub fn network_layer() -> Class {
     c
 }
 
+/// Textline (`0x0805`) — a segmented text line (box + baseline), the unit the
+/// LSTM line recognizer consumes (plan Phases 1B + 3E).
+#[must_use]
+pub fn textline() -> Class {
+    let mut c = Class::new("Textline");
+    c.language = Language::Unknown;
+    c.canonical_concept = Some("textline".to_string());
+    let mut baseline_y = Attribute::new("baseline_y");
+    baseline_y.type_name = Some("i32".to_string());
+    c.attributes = vec![baseline_y];
+    c.associations = vec![family_edge("belongs_to", "PageLayout")];
+    c
+}
+
+/// Blob (`0x0806`) — a connected-component blob container (Tesseract
+/// `C_BLOB`/`BLOBNBOX` kind), the page-segmentation substrate (plan Phase 3B/3D).
+#[must_use]
+pub fn blob() -> Class {
+    let mut c = Class::new("Blob");
+    c.language = Language::Unknown;
+    c.canonical_concept = Some("blob".to_string());
+    c.associations = vec![family_edge("belongs_to", "Textline")];
+    c
+}
+
+/// PageLayout (`0x0807`) — the layout-analysis result container: blocks /
+/// columns / reading order over one page (plan Phase 3F/4A).
+#[must_use]
+pub fn page_layout() -> Class {
+    let mut c = Class::new("PageLayout");
+    c.language = Language::Unknown;
+    c.canonical_concept = Some("page_layout".to_string());
+    c.associations = vec![family_edge("derived_from", "PageImage")];
+    c
+}
+
+/// PageImage (`0x0808`) — a rasterized page container (decoded + grey/binary
+/// normalized), the OCR input kind (plan Phase 2).
+#[must_use]
+pub fn page_image() -> Class {
+    let mut c = Class::new("PageImage");
+    c.language = Language::Unknown;
+    c.canonical_concept = Some("page_image".to_string());
+    let mut depth = Attribute::new("depth");
+    depth.type_name = Some("u8".to_string());
+    c.attributes = vec![depth];
+    c
+}
+
+/// OcrRenderer (`0x0809`) — an output-rendering KIND (txt / tsv / hOCR /
+/// searchable-PDF), one slot; the concrete format is the classid custom-low
+/// half, mirroring the `network_layer` pattern (plan Phase 4B-D).
+#[must_use]
+pub fn ocr_renderer() -> Class {
+    let mut c = Class::new("OcrRenderer");
+    c.language = Language::Unknown;
+    c.canonical_concept = Some("ocr_renderer".to_string());
+    let mut format = Attribute::new("format");
+    format.type_name = Some("u8".to_string());
+    c.attributes = vec![format];
+    c
+}
+
 // ─────────────────────────────────────────────────────────────────────
 // 0x09XX — Health domain (OGIT Healthcare). The reusable Active-Record
 // shape for the clinical concepts. `diagnosis` (0x0902) is the worked
@@ -5258,17 +5358,28 @@ mod tests {
         }
         // Counts line up with the codebook blocks.
         assert_eq!(concepts_in_domain(ConceptDomain::Health).count(), 7);
-        // 0x08XX OCR: the four container KINDS (unicharset/recoder/charset/
-        // network_layer). Content (the 112 unichars, code tables) never becomes
-        // concepts — see the CODEBOOK 0x08XX section note (Osint zero-rows
-        // precedent).
-        assert_eq!(concepts_in_domain(ConceptDomain::Ocr).count(), 4);
+        // 0x08XX OCR: the nine container KINDS (unicharset/recoder/charset/
+        // network_layer + the PDF→text plan mints textline/blob/page_layout/
+        // page_image/ocr_renderer). Content (the 112 unichars, code tables,
+        // blob outlines) never becomes concepts — see the CODEBOOK 0x08XX
+        // section note (Osint zero-rows precedent).
+        assert_eq!(concepts_in_domain(ConceptDomain::Ocr).count(), 9);
         let ocr: Vec<&str> = concepts_in_domain(ConceptDomain::Ocr)
             .map(|(name, _)| name)
             .collect();
         assert_eq!(
             ocr,
-            ["unicharset", "recoder", "charset", "network_layer"],
+            [
+                "unicharset",
+                "recoder",
+                "charset",
+                "network_layer",
+                "textline",
+                "blob",
+                "page_layout",
+                "page_image",
+                "ocr_renderer",
+            ],
             "OCR domain set drift — re-sync the consumer coverage gate",
         );
         assert_eq!(concepts_in_domain(ConceptDomain::HR).count(), 4);