Show which rewrite backends are available, which rule sources are loaded, and supported file types.

Return the editing rules (default rules, any installed humanizer skill, and the user's own SOP files).

Read these before rewriting text yourself. Custom rules outrank the defaults.

Read a PDF, DOCX, Markdown or text file and return its text as Markdown.

Args: path: Absolute path to the file on this machine. max_chars: Cap on returned text.

Check text for AI-writing tells without using a model.

Returns a 0-100 score (higher means more AI-like), counts by category, and each finding with its line, the matched text, a message and a suggested fix. Quotations are skipped except for chatbot artefacts.

Compare an original and a rewrite: AI-tell score before and after, remaining tells, and whether numbers, links, emails and quotations survived. Call this after rewriting, before saving.

Write rewritten Markdown text to a .docx, .pdf, .md or .txt file.

Args: text: The rewritten document as Markdown (headings with #, one blank line between paragraphs). output_path: Absolute path ending in .docx, .pdf, .md or .txt. source_path: Optional original file. If it is a DOCX with the same structure, its styles are kept. overwrite: Replace an existing output file. The source file is never overwritten.

Rewrite text on the server with a model backend, then lint it and check facts.

Args: text: Text to rewrite (Markdown or plain). level: light, standard or deep. backend: auto, claude-cli, anthropic or mock. Default auto. locale: For example "British English". style_sample: A sample of the author's own writing to match. assess: Also run the AI assessor for a before and after score (slower). voice: Match the writing samples in the voice folder (see humanizer_voice_profile). semantic: Add an AI pass that catches meaning-level tells such as empty claims (slower). local_model: Model name when backend is ollama or openai-compatible, for example llama3.1:8b.

Rewrite a PDF, DOCX, Markdown or text file on the server and save the result next to it.

PDF input is saved as .docx unless output_format is pdf. The original file is never changed. Default output is .humanised..

Args: path: Absolute path to the source file. output_path: Optional absolute output path (.docx, .pdf, .md or .txt). output_format: docx, pdf, md or txt. Default follows the source (PDF sources become docx). level: light, standard or deep. backend: auto, claude-cli, anthropic or mock. locale: For example "British English". style_sample: A sample of the author's own writing to match. overwrite: Replace an existing output file. assess: Also run the AI assessor for a before and after score (slower). also_pdf: Also save a PDF copy next to the output. voice: Match the writing samples in the voice folder. semantic: Add an AI pass that catches meaning-level tells (slower). local_model: Model name for the ollama or openai-compatible backends.

Score how AI-written a text reads (0 to 100, higher is more AI-like) with a blind rubric.

Returns six dimension scores, quoted issues that were checked against the text, and the pattern-checker score for comparison. Uses a model backend, so it is slower than lint_text.

Args: text: The text to assess. backend: auto, claude-cli, anthropic or mock. runs: Repeat the assessment and average it (reduces noise).

Score a PDF, DOCX, Markdown or text file the same way as humanizer_assess_text.

Blind before and after scores for an original and your rewrite, plus a faithfulness check.

Use this after rewriting text yourself. The assessor is not told which version is the rewrite. The faithfulness check reports details the rewrite invented, lost or changed.

Measure the author's writing habits from the samples in the voice folder.

Returns the folder path, files read, word count, measured habits (sentence length, contractions, dashes) and a short model-written description of the voice. At least 150 words of samples are needed.

Learn the author's taste from an AI draft and the version they fixed by hand.

Returns the sentences that changed, phrases they removed, and general rules. With save=true the phrases are added to the pattern checker and the rules to the rewrite prompt from then on.

Remove a phrase that was learned from the author's edits.