Extract pdf data to json

frappe_pdf_extractor/__init__.py

@@ -0,0 +1,7 @@
+__all__ = [
+    "extract_pdf_to_json",
+    "process_one",
+    "process_dir",
+]
+
+__version__ = "0.1.0"

frappe_pdf_extractor/api.py

@@ -0,0 +1,88 @@
+from __future__ import annotations
+
+from typing import Dict, List, Optional
+
+try:
+    import frappe  # type: ignore
+    from frappe.utils import get_site_path
+    from frappe.utils.file_manager import get_file_path  # type: ignore
+except Exception:  # pragma: no cover - allows import outside Frappe
+    frappe = None  # type: ignore
+    def get_site_path(*parts):  # type: ignore
+        import os
+        return os.path.abspath("/".join(parts))
+    def get_file_path(file_url: str):  # type: ignore
+        return file_url
+
+from .pdf_to_json import process_dir, process_one
+
+
+def _resolve_path_maybe_frappe(path_or_ref: str) -> str:
+    """Resolve a path that may be a filesystem path, a /files URL, or a File docname."""
+    import os
+
+    candidate = path_or_ref
+    # Absolute/relative path on filesystem
+    if os.path.exists(candidate):
+        return os.path.abspath(candidate)
+
+    # Site files or private files
+    if candidate.startswith("/files/") or candidate.startswith("/private/files/"):
+        return get_site_path(candidate.lstrip("/"))
+
+    # Try Frappe File doctype lookup
+    if frappe is not None:
+        try:
+            file_doc = frappe.get_doc("File", candidate)
+            file_url = getattr(file_doc, "file_url", None) or getattr(file_doc, "file_name", None)
+            if file_url:
+                return get_file_path(file_url)
+        except Exception:
+            pass
+
+    # As a last resort, return as-is
+    return candidate
+
+
+# Whitelisted API for Frappe/ERPNext
+if frappe is not None:
+    @frappe.whitelist(allow_guest=False)
+    def extract_pdf(pdf: Optional[str] = None, dir: Optional[str] = None, outdir: Optional[str] = None, keep_txt: int | bool = 0):
+        """Extract PDF(s) to JSON.
+
+        Args:
+            pdf: Path, /files URL, or File docname for a single PDF
+            dir: Path or /files directory containing PDFs (recursive)
+            outdir: Output directory (defaults to the same as the PDF or directory)
+            keep_txt: 1/0 to retain intermediate pdftotext output
+        Returns:
+            Dict with paths of generated JSON files
+        """
+        if not pdf and not dir:
+            frappe.throw("Provide either 'pdf' or 'dir'")
+
+        keep = bool(int(keep_txt)) if isinstance(keep_txt, (str, int)) else bool(keep_txt)
+
+        if pdf:
+            pdf_path = _resolve_path_maybe_frappe(pdf)
+            json_path = process_one(pdf_path, outdir=outdir, keep_txt=keep)
+            return {"json_path": json_path}
+        else:
+            dir_path = _resolve_path_maybe_frappe(dir)  # type: ignore[arg-type]
+            outputs = process_dir(dir_path, outdir=outdir, keep_txt=keep)
+            return {"json_paths": outputs}
+
+else:
+    # Non-Frappe shim for testing
+    def extract_pdf(pdf: Optional[str] = None, dir: Optional[str] = None, outdir: Optional[str] = None, keep_txt: int | bool = 0):
+        if not pdf and not dir:
+            raise ValueError("Provide either 'pdf' or 'dir'")
+        keep = bool(int(keep_txt)) if isinstance(keep_txt, (str, int)) else bool(keep_txt)
+        if pdf:
+            pdf_path = _resolve_path_maybe_frappe(pdf)
+            json_path = process_one(pdf_path, outdir=outdir, keep_txt=keep)
+            return {"json_path": json_path}
+        else:
+            dir_path = _resolve_path_maybe_frappe(dir)  # type: ignore[arg-type]
+            outputs = process_dir(dir_path, outdir=outdir, keep_txt=keep)
+            return {"json_paths": outputs}

frappe_pdf_extractor/hooks.py

@@ -0,0 +1,8 @@
+app_name = "frappe_pdf_extractor"
+app_title = "Frappe PDF Extractor"
+app_publisher = "Your Company"
+app_description = "Universal PDF to JSON extraction via pdftotext"
+app_email = "dev@example.com"
+app_version = "0.1.0"
+
+# Whitelist methods are declared in api.py with @frappe.whitelist

frappe_pdf_extractor/modules.txt

@@ -0,0 +1 @@
+frappe_pdf_extractor