OrgLens · OpenAI Edition

A standalone, server-backed reorganization analysis prototype. Python Flask backend; OpenAI Responses API; Russian dark interface. Import before/after TXT, MD, DOCX, PDF (with extractable text), XLSX or paste text. Get structured evidence-grounded findings, PDF via browser print, JSON export and analyst chat.

MacBook / Windows startup

1. Install Python 3.11+.
2. Open Terminal in this folder. On Mac, use python3 -m venv .venv && source .venv/bin/activate; on Windows use py -m venv .venv && .venv\\Scripts\\activate.
3. python -m pip install -r requirements.txt
4. Copy .env.example to .env; replace the placeholder with a NEW OpenAI key. Never reuse a key shared in chat. On Mac: cp .env.example .env.
5. python app.py then open http://127.0.0.1:5000.
6. Run python -m pytest -q for offline checks.

Hosting

A static-only chatgpt.site deployment cannot run Python or protect a key itself. Deploy Flask to a backend that supports Python and environment secrets. Configure OPENAI_API_KEY and ORGLENS_ACCESS_TOKEN on the host and proxy both frontend and /api/* to Flask, or connect the static frontend to an HTTPS Flask endpoint (requires updating fetch paths and CORS). Never publish .env. Do not expose upload endpoints publicly without authentication, throttling, privacy review and cost caps; optional token gate in this prototype is a basic demo gate, not production authentication. Flask's built-in server is for local testing, not production.

Privacy and limitations

Original documents are parsed in memory; extracted text is sent to OpenAI with store=False after consent. No persistent file storage is added. The original local-only privacy claim is no longer true. No scanned PDF OCR. Long files are capped at 110 segments / 40k chars per side, with a visible warning; PDFs limited to 75 pages; XLSX to 12 sheets x 1500 rows. Automated analysis is a review aid, not definitive legal or organizational advice. Quotes and locators are from parser-extracted source segments and model-made findings without valid evidence IDs are dropped. Authentication, rate limiting, production file scanning and further access controls are required for external deployment.

Double-click launcher: start_mac.command (Mac), start_windows.bat (Windows). macOS may require granting permission to a downloaded script in Privacy & Security. The initial Mac launch creates .env for editing; launch again after adding your new key.