Show HN: DocSumm AI – Source-linked summaries for long PDFs/URLs (github.com)

DocSumm AI is an open-source tool that produces opinionated, source-linked one-line summaries of long PDFs, Word docs, or text files with an emphasis on meaning retention rather than fitting token limits. Built for researchers, analysts, and AI developers, it uses context-aware chunking (semantic segmentation) and adaptive compression to preserve important insights across sections, outputs human- and machine-readable JSON/Markdown, and is accessible via a CLI and Python API for easy pipeline integration. Why this matters: many summarizers naively truncate context to meet token constraints and produce shallow or inaccurate outputs; DocSumm AI addresses that by dynamically segmenting documents and compressing each part to keep salient information, enabling higher-fidelity summaries that remain concise. Technical implications include reproducible, scriptable workflows (summarize() in Python or docsumm CLI), multi-format support (PDF, DOCX, TXT), and transparent outputs for downstream analysis or auditing. Released under an MIT license, it’s positioned for community collaboration and integration into ML research and production pipelines that need reliable, source-aware document condensation.