pdf-slice: A Local PDF Extractor That Flags Its Own Failures Instead of Hiding Them
The developer of pdf-slice, a small CPU-only PDF-to-markdown tool, built it after repeated silent extraction failures in existing pipelines. The tool keys its cache by full URL and a SHA-256 hash of the file bytes rather than filename, after a bug served an April report in place of a March one. When native text extraction looks suspiciously thin, such as a 46-page file yielding only about 3,000 characters, it automatically falls back to OCR and labels low-confidence results with a warning. Each page carries provenance metadata, and encrypted, corrupt, or empty inputs raise clear errors rather than producing empty output. The project also ships an MCP server with extract and cross-route verification tools, and its bundled test matrix reports zero silent errors across its fixtures.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.
Discussion (0)
Log in to join the discussion and vote.
Log in