add doctor command, fix ISBN false positives, wire up CI

ISBN regex now requires an explicit prefix, so it stops matching
years, URLs, job IDs, and dilution ratios.

doctor checks the environment end to end: Python, Tesseract,
PaddleOCR, the PyTorch backend, table/figure detectors, scispaCy,
config, DB, output.

CI runs on 2 OS x 3 Python versions plus a doctor smoke job.
Also added a changelog

src/ocr_pipeline/doctor.py

@@ -0,0 +1,191 @@
+from __future__ import annotations
+
+import importlib.util
+import shutil
+import sys
+from dataclasses import dataclass
+from typing import Any
+
+from ocr_pipeline.config import settings
+from ocr_pipeline.onboarding import tesseract_install_hint
+from ocr_pipeline.utils.logging import get_logger
+
+logger = get_logger(__name__)
+
+
+@dataclass(frozen=True)
+class CheckResult:
+    name: str
+    status: str  # "ok" | "warn" | "fail"
+    detail: str
+    hint: str | None = None
+
+
+def _has_module(name: str) -> bool:
+    return importlib.util.find_spec(name) is not None
+
+
+def check_tesseract() -> CheckResult:
+    path = shutil.which("tesseract")
+    if path:
+        return CheckResult("Tesseract", "ok", f"found at {path}")
+    return CheckResult("Tesseract", "fail", "not on PATH", f"Install: {tesseract_install_hint()}")
+
+
+def check_paddle() -> CheckResult:
+    if not _has_module("paddleocr"):
+        return CheckResult("PaddleOCR", "warn", "not installed", "uv sync --extra paddle")
+    try:
+        from paddleocr import PaddleOCR  # type: ignore[import-not-found]
+
+        PaddleOCR(use_angle_cls=True, lang="en", use_gpu=False, show_log=False)
+        return CheckResult("PaddleOCR", "ok", "initialized (en)")
+    except Exception as exc:
+        return CheckResult("PaddleOCR", "warn", f"init failed: {exc}")
+
+
+def check_torch() -> CheckResult:
+    if not _has_module("torch"):
+        return CheckResult("PyTorch", "warn", "not installed", "uv sync --extra tables")
+    import torch  # type: ignore[import-not-found]
+
+    cuda = torch.cuda.is_available()
+    mps = getattr(torch.backends, "mps", None) is not None and torch.backends.mps.is_available()
+    backend = "cuda" if cuda else "mps" if mps else "cpu"
+    return CheckResult("PyTorch", "ok", f"backend={backend}")
+
+
+def check_tables() -> CheckResult:
+    if not _has_module("transformers") or not _has_module("torch"):
+        return CheckResult(
+            "Table detector", "warn", "missing torch/transformers", "uv sync --extra tables"
+        )
+    return CheckResult("Table detector", "ok", "transformers + torch present")
+
+
+def check_figures() -> CheckResult:
+    if not _has_module("layoutparser"):
+        return CheckResult(
+            "Figure detector", "warn", "layoutparser not installed", "uv sync --extra figures"
+        )
+    try:
+        import layoutparser as lp  # type: ignore[import-not-found]
+
+        has_detectron = hasattr(lp, "Detectron2LayoutModel")
+        has_auto = hasattr(lp, "AutoLayoutModel")
+        if not (has_detectron or has_auto):
+            return CheckResult(
+                "Figure detector",
+                "warn",
+                "no layout model backend",
+                "Install Detectron2 for your platform",
+            )
+        return CheckResult("Figure detector", "ok", "layoutparser + backend present")
+    except Exception as exc:
+        return CheckResult("Figure detector", "warn", f"import failed: {exc}")
+
+
+def check_spacy() -> CheckResult:
+    if not _has_module("spacy"):
+        return CheckResult(
+            "scispaCy NER", "warn", "spacy not installed", "uv sync --extra scientific"
+        )
+    import spacy  # type: ignore[import-not-found]
+
+    try:
+        spacy.load(settings.entities.model)
+        return CheckResult("scispaCy NER", "ok", f"loaded {settings.entities.model}")
+    except OSError:
+        return CheckResult(
+            "scispaCy NER",
+            "warn",
+            f"model {settings.entities.model!r} not downloaded",
+            "uv pip install https://s3-us-west-2.amazonaws.com/ai2-s2-scispacy/releases/v0.5.4/en_core_sci_lg-0.5.4.tar.gz",
+        )
+
+
+def check_config() -> CheckResult:
+    from ocr_pipeline.config import find_config_file
+
+    path = find_config_file()
+    if path is None:
+        return CheckResult("Config", "warn", "no config.yaml found", "Run: ocr-pipeline init")
+    return CheckResult("Config", "ok", f"loaded from {path}")
+
+
+def check_db() -> CheckResult:
+    db_path = settings.watch_db_path
+    try:
+        db_path.parent.mkdir(parents=True, exist_ok=True)
+        with db_path.open("a"):
+            pass
+        return CheckResult("Database", "ok", f"writable at {db_path}")
+    except OSError as exc:
+        return CheckResult("Database", "fail", f"not writable: {exc}")
+
+
+def check_output_dir() -> CheckResult:
+    out = settings.output_dir
+    try:
+        out.mkdir(parents=True, exist_ok=True)
+        return CheckResult("Output dir", "ok", str(out))
+    except OSError as exc:
+        return CheckResult("Output dir", "fail", f"cannot create: {exc}")
+
+
+def check_python() -> CheckResult:
+    return CheckResult(
+        "Python",
+        "ok",
+        f"{sys.version_info.major}.{sys.version_info.minor}.{sys.version_info.micro}",
+    )
+
+
+CHECKS = [
+    check_python,
+    check_tesseract,
+    check_paddle,
+    check_torch,
+    check_tables,
+    check_figures,
+    check_spacy,
+    check_config,
+    check_db,
+    check_output_dir,
+]
+
+
+def run_checks() -> list[CheckResult]:
+    results: list[CheckResult] = []
+    for check in CHECKS:
+        try:
+            results.append(check())
+        except Exception as exc:  # noqa: BLE001 - surfaced as a fail result, never propagated
+            results.append(CheckResult(check.__name__, "fail", f"unexpected error: {exc}"))
+    return results
+
+
+def render(results: list[CheckResult], console: Any) -> None:
+    from rich.table import Table
+
+    status_style = {"ok": "green", "warn": "yellow", "fail": "red"}
+    icon = {"ok": "[green]OK[/green]", "warn": "[yellow]WARN[/yellow]", "fail": "[red]FAIL[/red]"}
+    table = Table(title="ocr-pipeline health check")
+    table.add_column("Status")
+    table.add_column("Check", style="cyan")
+    table.add_column("Detail")
+    for r in results:
+        table.add_row(icon[r.status], r.name, r.detail)
+    console.print(table)
+
+    failing = [r for r in results if r.status == "fail"]
+    warning = [r for r in results if r.status == "warn"]
+    if failing or warning:
+        console.print()
+        for r in failing + warning:
+            if r.hint:
+                console.print(f"  [{status_style[r.status]}]{r.name}:[/] {r.hint}")
+
+
+def exit_code_for(results: list[CheckResult]) -> int:
+    return 1 if any(r.status == "fail" for r in results) else 0