Personal AI, On Personal Devices
Python
"""PDF text extraction tool — extract text from PDF files via pdfplumber."""
from __future__ import annotations
from pathlib import Path
from typing import Any, List
from openjarvis.core.registry import ToolRegistry
from openjarvis.core.types import ToolResult
from openjarvis.tools._stubs import BaseTool, ToolSpec
_DEFAULT_MAX_CHARS = 50_000
def _parse_pages(pages_str: str, total_pages: int) -> List[int]:
"""Parse a page range string into zero-indexed page numbers.
Supports formats like ``"1-5"`` (range) and ``"1,3,5"`` (list).
Page numbers in the input are 1-indexed; the returned list is 0-indexed.
Parameters
----------
pages_str:
Page specification string.
total_pages:
Total number of pages in the PDF.
Returns
-------
List[int]
Sorted list of zero-indexed page numbers.
"""
result: list[int] = []
for part in pages_str.split(","):
part = part.strip()
if not part:
continue
if "-" in part:
start_str, end_str = part.split("-", 1)
start = max(1, int(start_str.strip()))
end = min(total_pages, int(end_str.strip()))
result.extend(range(start - 1, end))
else:
page_num = int(part)
if 1 <= page_num <= total_pages:
result.append(page_num - 1)
return sorted(set(result))
@ToolRegistry.register("pdf_extract")
class PDFExtractTool(BaseTool):
"""Extract text content from PDF files using pdfplumber."""
tool_id = "pdf_extract"
@property
def spec(self) -> ToolSpec:
return ToolSpec(
name="pdf_extract",
description=(
"Extract text from a PDF file. Returns the extracted text content."
),
parameters={
"type": "object",
"properties": {
"file_path": {
"type": "string",
"description": "Path to the PDF file.",
},
"pages": {
"type": "string",
"description": (
"Page range to extract, e.g. '1-5' or '1,3,5'."
" Omit to extract all pages."
),
},
"max_chars": {
"type": "integer",
"description": ("Maximum characters to return. Default 50000."),
},
},
"required": ["file_path"],
},
category="media",
required_capabilities=["file:read"],
)
def execute(self, **params: Any) -> ToolResult:
file_path = params.get("file_path", "")
if not file_path:
return ToolResult(
tool_name="pdf_extract",
content="No file_path provided.",
success=False,
)
path = Path(file_path)
# Validate extension
if path.suffix.lower() != ".pdf":
return ToolResult(
tool_name="pdf_extract",
content=f"Not a PDF file: {file_path}",
success=False,
)
# Check sensitive file policy
from openjarvis.security.file_policy import is_sensitive_file
if is_sensitive_file(path):
return ToolResult(
tool_name="pdf_extract",
content=f"Access denied: {file_path} is a sensitive file.",
success=False,
)
if not path.exists():
return ToolResult(
tool_name="pdf_extract",
content=f"File not found: {file_path}",
success=False,
)
try:
import pdfplumber
except ImportError:
return ToolResult(
tool_name="pdf_extract",
content=(
"pdfplumber package not installed."
" Install with: pip install pdfplumber"
),
success=False,
)
max_chars = params.get("max_chars", _DEFAULT_MAX_CHARS)
pages_param = params.get("pages")
try:
with pdfplumber.open(str(path)) as pdf:
total_pages = len(pdf.pages)
if pages_param:
page_indices = _parse_pages(pages_param, total_pages)
else:
page_indices = list(range(total_pages))
text_parts: list[str] = []
for idx in page_indices:
if 0 <= idx < total_pages:
page_text = pdf.pages[idx].extract_text() or ""
text_parts.append(page_text)
text = "\n\n".join(text_parts)
if len(text) > max_chars:
text = text[:max_chars] + "\n\n[Content truncated]"
return ToolResult(
tool_name="pdf_extract",
content=text or "No text content found in PDF.",
success=True,
metadata={
"file_path": str(path.resolve()),
"total_pages": total_pages,
"pages_extracted": len(page_indices),
},
)
except Exception as exc:
return ToolResult(
tool_name="pdf_extract",
content=f"PDF extraction error: {exc}",
success=False,
)
__all__ = ["PDFExtractTool"]