dl4rce/flaiwheel

What's New

Flaiwheel now indexes 9 file formats — not just markdown.

Supported Input Formats

| Format | Extension(s) | How it works |
|--------|-------------|--------------|
| Markdown | .md | Native (pass-through) |
| Plain text | .txt | Wrapped in heading |
| PDF | .pdf | Text extracted per page via pypdf |
| HTML | .html, .htm | Headings/lists/code converted to markdown, scripts stripped |
| reStructuredText | .rst | Heading underlines converted, code blocks preserved |
| Word | .docx | Paragraphs + heading styles mapped to markdown |
| JSON | .json | Pretty-printed in fenced code block |
| YAML | .yaml, .yml | Wrapped in fenced code block |
| CSV | .csv | Converted to markdown table |

Key Design

• All formats converted to markdown-like text in memory — no files generated on disk
• Original files stay untouched in the knowledge repo
• Feeds into the existing chunker pipeline (heading-based, fixed, or hybrid)
• Quality checks apply only to native .md files; other formats are indexed as-is
• If an optional dep is missing, the file is skipped gracefully (no crash)
• Hash-based diff caching works across all formats

New Dependencies

• pypdf (PDF extraction, pure Python)
• beautifulsoup4 (HTML parsing, pure Python)
• python-docx (DOCX extraction, pure Python)

Stats

• 155 tests (29 new reader tests)
• 3 new pure-Python dependencies (~3MB total)

Upgrade

Run the install script from your project directory to update.