AllRead

Read any file as clean text in one line — local, private, encoding-safe.

import allread

text = allread.read("report.csv")   # markdown table
text = allread.read("notes.txt")    # plain text
text = allread.read("page.html")    # extracted text + tables

No config. No encoding crashes. Works the same on Windows, macOS and Linux.

Install

pip install allread            # core: txt, md, rst, log, csv, tsv, json, jsonl, xml, html
pip install "allread[pdf]"     # + PDF support (PyMuPDF)
pip install "allread[docx]"    # + Word documents
pip install "allread[xlsx]"    # + Excel workbooks
pip install "allread[all]"     # everything

Optional formats are lazy: allread.read("report.pdf") raises a clear ImportError telling you exactly which extra to install.

Why AllRead?

• Encoding-safe by default — UTF-8, UTF-16/32, UTF-8-BOM, cp1252 and legacy encodings are handled automatically; Persian/Arabic/any non-ASCII content just works.
• Zero required dependencies — the core is pure standard library. Heavy engines are optional extras, loaded lazily.
• Structured files become Markdown — CSV/TSV/Excel come back as Markdown tables, ready for LLM prompts, notebooks or docs.
• Same API everywhere — read() returns text, load() returns a Document with source and format metadata.

API

Examples

doc = allread.load("sales.xlsx")
print(doc.format)        # "xlsx"
print(doc.text)          # markdown tables, one section per sheet

Roadmap

• Core formats (txt, md, rst, log, csv, tsv, json, jsonl, xml, html)
• Optional PDF / DOCX / XLSX engines with lazy imports
• OCR for scanned PDFs and images (allread[ocr])
• Audio transcription (allread[audio])
• CLI: allread file.pdf > out.md

License

MIT