A neutral prose naturalization tool. Paste English prose; Rewrite-Rescue diagnoses linguistic patterns that make it read as generated — semantic smoothing, evaluative inflation, formulaic structures, monotone rhythm, generation artifacts — and can revise for specificity, rhythm, and register fit while preserving meaning.
- Diagnoses linguistic patterns and explains each finding.
- Revises with deletion, simplification, and restructuring — never invention.
- Asks you for missing specifics instead of fabricating them.
- Runs a Demo mode that needs no API credentials at all.
- It does not judge whether a text was written by a person or a model.
- It does not promise outcomes with any third-party analysis tools.
- It does not store your text, style samples, or API keys. Session only.
- It does not fabricate facts, examples, citations, or experience.
Build in progress: M1 (skeleton + shell) complete. M2 diagnostics, M3 provider layer, and M4 audits + full deployment docs follow. Complete install, provider, and deployment directions land with M4.
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
streamlit run app.py
Demo mode works immediately — no credentials required.
Never commit .env or .streamlit/secrets.toml. The public app needs no
secrets: Demo mode and session-entered credentials cover it.
.streamlit/secrets.example.toml documents the fields for private deployments.
MIT — see LICENSE.