Allow signaling issues in label verification

copienator/commands/annotating.py

@@ -64,7 +64,7 @@ def make_base_image(pdf_path):
 
     current_y = 0
     for page in pages:
-        base_img.paste(page.convert("RGBA"), (0, current_y))
+        base_img.paste(page.convert("RGBA"), ((max_w - page.width) // 2, current_y))
         current_y += page.height
     return (base_img, total_h, max_w)
 

copienator/commands/cutleft.py

@@ -471,7 +471,7 @@ def run(
 
 def build_parser() -> argparse.ArgumentParser:
     parser = target_parser("Interactively crop the label margin from PDF copies")
-    parser.add_argument("--marked", action="store_true", help="Review flagged copies and clear each flag after saving with Enter")
+    parser.add_argument("--marked", action="store_true", help="Review flagged copies and clear only margin-cropping flags after saving with Enter")
     parser.add_argument(
         "--fullpage",
         action="store_true",

copienator/commands/gemini_for_labels.py

@@ -16,6 +16,7 @@ from pydantic import BaseModel, Field
 
 from copienator import configuration as config
 from copienator.gemini_config import thinking_config
+from copienator.copy_errors import marked_copy_paths
 from copienator import (
     CliError,
     EvaluationWorkspace,
@@ -530,6 +531,7 @@ def run(
     targets: list[Path],
     *,
     overwrite: bool = False,
+    marked: bool = False,
     client=None,
     sleep: Sleep = time.sleep,
     max_workers: int = 12,
@@ -538,6 +540,12 @@ def run(
         raise CliError("Le nombre de requêtes simultanées doit être positif.",
                        ExitCode.INVALID_ARGUMENTS)
     workspace.require_files("labels")
+    if marked:
+        targets = marked_copy_paths(workspace)
+        if not targets:
+            print("No flagged copies to detect labels for.")
+            return ExitCode.SUCCESS
+        overwrite = True
     images, warnings = selected_images(workspace, targets)
     for warning in warnings:
         print(f"Warning: {warning}")
@@ -626,6 +634,8 @@ def build_parser() -> argparse.ArgumentParser:
         type=Path,
         help="Additional copy PDFs or Cutleft images from the same evaluation",
     )
+    parser.add_argument("--marked", action="store_true",
+                        help="Regenerate labels only for copies with remaining flags")
     parser.add_argument(
         "--overwrite",
         action="store_true",
@@ -664,7 +674,7 @@ def main(argv: Sequence[str] | None = None) -> int:
                     ExitCode.INVALID_ARGUMENTS,
                 )
             targets.append(resolved)
-        return run(workspace, targets, overwrite=args.overwrite, max_workers=args.workers)
+        return run(workspace, targets, overwrite=args.overwrite, marked=args.marked, max_workers=args.workers)
 
     return execute(parser, argv, handle)
 

copienator/commands/grouping.py

@@ -117,7 +117,7 @@ def group_files(file_list):
     return [g['items'] for g in groups]
 
 def stitch_pdf_pages(images_list):
-    """Vertically concatenates a list of PIL images with no separator."""
+    """Stack PIL images vertically, centering each within the widest image."""
     if not images_list:
         return None
     if len(images_list) == 1:
@@ -130,7 +130,7 @@ def stitch_pdf_pages(images_list):
 
     y_offset = 0
     for img in images_list:
-        combined.paste(img, (0, y_offset))
+        combined.paste(img, ((max_width - img.width) // 2, y_offset))
         y_offset += img.height
 
     return combined

copienator/commands/modify_annotations.py

@@ -0,0 +1,73 @@
+from __future__ import annotations
+
+from collections.abc import Sequence
+from pathlib import Path
+from tempfile import TemporaryDirectory
+
+from copienator import (
+    EvaluationWorkspace,
+    ExitCode,
+    evaluation_parser,
+    execute,
+    read_json,
+    workspace_from_args,
+)
+from copienator.annotation_modifications import (
+    APPLIED,
+    DRAFT,
+    MODES,
+    baseline,
+    output_removals,
+    read_instructions,
+    rebuild,
+    validate_instructions,
+)
+from copienator.filesystem import publish_file_sets
+
+
+def run(workspace: EvaluationWorkspace, *, annotation_dir: str = "BGnot") -> ExitCode:
+    workspace.require_files("labels", DRAFT)
+    workspace.require_directories("Copies", annotation_dir)
+    text = (workspace.root / DRAFT).read_text(encoding="utf-8")
+    instructions = read_instructions(text, annotation_dir)
+    validate_instructions(workspace.root, annotation_dir, instructions)
+    directory = workspace.root / annotation_dir
+    applied = directory / APPLIED
+    previous = read_instructions(applied.read_text(encoding="utf-8"), annotation_dir) if applied.exists() else []
+    copy_ids = sorted({item.copy_id for item in [*previous, *instructions]})
+    # Prepare every copy before publishing anything, including the applied file.
+    with TemporaryDirectory(prefix=".annotation-edits-", dir=directory) as temporary:
+        staging_root = Path(temporary)
+        file_sets = {}
+        for copy_id in copy_ids:
+            output = directory / f"Copie{copy_id}"
+            meta = baseline(output)
+            staging = staging_root / output.name
+            staging.mkdir()
+            selected = [item for item in instructions if item.copy_id == copy_id]
+            rebuild(workspace.root, annotation_dir, output, staging, meta, selected,
+                    read_json(output / "score.json"))
+            file_sets[output] = (staging, output_removals(output))
+        manifest = staging_root / "manifest"
+        manifest.mkdir()
+        (manifest / APPLIED).write_text(text, encoding="utf-8")
+        file_sets[directory] = (manifest, ())
+        publish_file_sets(file_sets)
+    print(f"Modifications appliquées à {len(copy_ids)} copie(s). Les scores sont conservés.")
+    print("Relancez « Attribuer les noms et préparer A Rendre » pour actualiser tous les livrables.")
+    return ExitCode.SUCCESS
+
+
+def build_parser():
+    parser = evaluation_parser("Modify reviewed answer bodies using annotation_modifications.txt")
+    parser.add_argument("--annotation-dir", choices=MODES, default="BGnot")
+    return parser
+
+
+def main(argv: Sequence[str] | None = None) -> int:
+    return execute(build_parser(), argv, lambda args: run(
+        workspace_from_args(args), annotation_dir=args.annotation_dir))
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

copienator/commands/plotting.py

@@ -21,6 +21,7 @@ from copienator import (
     target_parser,
     workspace_from_target,
 )
+from copienator.copy_errors import clear_copy_error, mark_copy_error, marked_copy_paths
 from copienator.platform import open_path
 from copienator.utils import natural_key, read_all_labels
 
@@ -307,6 +308,7 @@ class ImageViewer:
         self.forward_stack = []
         self.current_pil_image = None
         self.failed = False
+        self.skipped_copies = set()
 
         from config import PLOTTING_KB
 
@@ -314,6 +316,7 @@ class ImageViewer:
         self.root.bind(PLOTTING_KB["OK"], self.on_enter)
         self.root.bind(PLOTTING_KB["previous"], self.on_previous)
         self.root.bind("P", self.on_previous)
+        self.root.bind("!", self.on_skip)
         self.root.bind(PLOTTING_KB["edit"], self.on_edit)
         self.root.bind(PLOTTING_KB["open pdf"], self.on_open_pdf)
         self.root.bind(PLOTTING_KB["open original pdf"], self.on_open_ori_pdf)
@@ -345,6 +348,10 @@ class ImageViewer:
                     self.root.quit()
                     return
 
+                if metadata["copie"] in self.skipped_copies:
+                    self.root.after(0, self.poll_queue)
+                    return
+
                 # Check if we switched to a new "Copie" group
                 if self.active_copie_name != metadata["copie"]:
                     self.save_current_batch() # Write previous group to disk
@@ -361,8 +368,21 @@ class ImageViewer:
         """Writes the accumulated data to the main JSON file."""
         if self.active_copie_name and self.accumulated_results:
             main_json_path = self.base_dir / "Copies" / f"{self.active_copie_name}.json"
+            accepted = [item[2] for item in self.history
+                        if item[3] == self.active_copie_name]
+            parts = {item["part"] for item in accepted}
+            expected = (set(range(1, len(accepted[-1]["schema"]["columns_per_file"]) + 1))
+                        if accepted else set())
+            if not expected or parts != expected:
+                mark_copy_error(self.workspace, main_json_path.with_suffix(".pdf"),
+                                "Vérification des labels incomplète", source="plotting")
+                self.failed = True
+                self.accumulated_results = None
+                return
             print(f"Writing aggregated result to {main_json_path}")
             atomic_write_json(main_json_path, self.accumulated_results)
+            clear_copy_error(self.workspace, main_json_path.with_suffix(".pdf"),
+                             source="plotting")
             self.accumulated_results = None
 
     def close(self):
@@ -467,6 +487,21 @@ class ImageViewer:
             self.is_viewing = False
             self.label.config(image="", text="Loading next...")
 
+    def on_skip(self, _event=None):
+        if not self.is_viewing:
+            return
+        copie = self.current_meta["copie"]
+        mark_copy_error(self.workspace, self.workspace.copies_dir / f"{copie}.pdf",
+                        "Problème repéré pendant la vérification des labels", source="plotting")
+        self.failed = True
+        self.skipped_copies.add(copie)
+        # Discard accepted parts of this copy without replacing its previous result.
+        self.accumulated_results = None
+        self.history = [item for item in self.history if item[3] != copie]
+        self.forward_stack = [item for item in self.forward_stack if item[2]["copie"] != copie]
+        self.is_viewing = False
+        self.label.config(image="", text="Copie signalée. Chargement de la suivante…")
+
     def on_open_pdf(self, event):
         if self.is_viewing and self.current_json_path:
             a = self.current_json_path.stem.split('_')[0] + ".pdf"
@@ -530,17 +565,24 @@ def _selected_images(workspace: EvaluationWorkspace, target: Path) -> list[Path]
     return sorted(workspace.cutleft_dir.glob("*.jpg"), key=natural_key)
 
 
-def run(workspace: EvaluationWorkspace, target: Path) -> ExitCode:
+def run(workspace: EvaluationWorkspace, target: Path, *, marked: bool = False) -> ExitCode:
     workspace.require_files("labels")
     all_labels = read_all_labels(workspace.root)
-    files_to_process = _selected_images(workspace, target)
+    if marked:
+        copies = marked_copy_paths(workspace)
+        if not copies:
+            print("No flagged copies to verify.")
+            return ExitCode.SUCCESS
+        files_to_process = [image for pdf in copies for image in _selected_images(workspace, pdf)]
+    else:
+        files_to_process = _selected_images(workspace, target)
     if not files_to_process:
         print(f"No Cutleft images found for {target}")
         return ExitCode.PARTIAL
 
     print(
         "o to open pdf, O original pdf, e to edit part, p to go back, "
-        "i to open the statement, click for coordinates"
+        "i to open the statement, ! to flag and skip the copy, click for coordinates"
     )
     input_queue = queue.Queue(maxsize=5)
     worker = threading.Thread(
@@ -556,7 +598,9 @@ def run(workspace: EvaluationWorkspace, target: Path) -> ExitCode:
 
 
 def build_parser() -> argparse.ArgumentParser:
-    return target_parser("Interactively verify detected label coordinates")
+    parser = target_parser("Interactively verify detected label coordinates")
+    parser.add_argument("--marked", action="store_true", help="Verify only copies with remaining flags")
+    return parser
 
 
 def main(argv: Sequence[str] | None = None) -> int:
@@ -564,7 +608,7 @@ def main(argv: Sequence[str] | None = None) -> int:
 
     def handle(args: argparse.Namespace) -> ExitCode:
         workspace, target = workspace_from_target(args)
-        return run(workspace, target)
+        return run(workspace, target, marked=args.marked)
 
     return execute(parser, argv, handle)