Repository navigation
Privacy and AI
This page is generated from
docs/community/wiki/Privacy-and-AI.md. Edit the source in the repository and the change will be synced here.
ReboundOps is designed so that your job search stays on your computer unless you decide otherwise. This page summarises the behaviour described in the README and docs/RELEASE_READINESS.md.
- One SQLite database in
~/.reboundops, or the folder inREBOUND_DATA_DIR. The folder is created with0700permissions and the files with0600(best effort on Windows). - The server listens only on a loopback address (
127.0.0.1by default) and refuses other host names. - Logs default to
warnand never include request bodies or URLs. - There is no account, no telemetry and no project-run cloud service.
| Feature | Sent | When |
|---|---|---|
| Local only (default) | Nothing | Always |
| Discover | The board name you typed, to that job board's public HTTPS endpoint | Only after you allow it. One request per second per host, no cookies, 30-minute cache |
| Ollama | The text a task needs, to Ollama on your own machine | When you run an assist task. Ollama must be on a loopback address |
| Cloud AI (Anthropic, OpenAI, Gemini, OpenRouter) | Only the text a task needs, using verified profile entries only | Only with your consent for that request. Keys come from environment variables and are never stored or shown |
Cloud providers are tested with mocked requests only. Their own privacy terms apply to anything you send them.
AI only suggests: rewording, summaries, missing skills, interview questions and drafts. Each suggestion is checked against your records, and anything that introduces a fact you have not recorded is withheld. Nothing an AI suggests is saved until you choose to save it. Every feature works with AI off.
Backups can be password-protected (scrypt and AES-256-GCM). The password is never stored and cannot be recovered. Keep backups somewhere you control.
Stop ReboundOps and delete the data folder. Download a backup first if you might want the data later.