This project is pre-1.0. Only the latest commit on the default branch is supported. There are no maintained release branches yet.

Do not open a public issue for security problems.

If you've found a vulnerability, please email:

boyciscox@gmail.com

Include as much as you can:

• A clear description of the issue
• Steps to reproduce
• The potential impact (data exposure, credential leak, remote code execution, denial of service, etc.)
• Any proof-of-concept code or terminal output
• The version/commit you tested against
• Whether you want public credit once it's fixed

You'll get an acknowledgement as soon as possible. If you don't hear back within a few days, feel free to send a follow-up.

Please give us a reasonable amount of time to fix the issue before disclosing it publicly. We're a small project and we'd rather ship a fix than a press release.

• Credential or API key exposure
• Injection vulnerabilities (SQL, prompt injection that leads to unwanted external effects, etc.)
• Path traversal or arbitrary file read/write
• Denial of service that affects other users
• Cross-user data leakage (a user reading another user's research)
• Dependency vulnerabilities that are exploitable in the default configuration

• The AI producing factually wrong answers (that's a research-quality issue — open a normal issue using the research quality template)
• DuckDuckGo rate limiting you
• Gemini rate limiting you
• Free-tier quota exhaustion
• The bot not being reachable when your computer is off

This project stores credentials in .env, which is in .gitignore. If you have ever committed .env by accident:

1. Rotate the affected keys immediately (Telegram bot token via @BotFather, Gemini key via Google AI Studio).
2. Remove the file from Git history using git filter-repo or BFG Repo-Cleaner.
3. Force-push the cleaned history.

Deleting .env in a new commit is not enough — the file is still in the history and can be recovered.

This is a self-hosted research assistant. There is no hosted service, no shared infrastructure, and no user account system beyond the Telegram user ID. The main threats are:

• Leaking your own API keys
• A malicious web page causing the fetcher to do something unexpected
• Prompt injection from source content steering the writer

The last one is a known, accepted limitation. The pipeline constrains the writer to the research package, but a sufficiently adversarial source could still bias the model's output. Fact-checking against the package reduces but does not eliminate this risk.