This commit completes the "Strictly Local" journey by adding several key features:
- Permanent `security.strictLocal` configuration key.
- Hardened Docker sandbox isolation (enforces `network: "none"`).
- Local compliance checks in `moltbot security audit`.
- Ollama health and model checks in `moltbot doctor`.
- Local CLI-based TTS support (offline speech).
- Skill vetting to prevent external URL leakage in strict mode.
- Documentation and unit tests for all new functionality.
Co-authored-by: samibs <1743891+samibs@users.noreply.github.com>
This commit introduces a new `--local-only` flag for the `gateway` command that hardens the configuration for strictly local usage by:
- Disabling update checks on start
- Disabling diagnostics
- Clearing model fallbacks (to prevent accidental cloud usage)
- Denying external tools (web_search, web_fetch, browser)
- Enabling Docker sandboxing for non-main sessions
It also adds a detailed `SECURITY_ASSESSMENT_LOCAL.md` report and a unit test for the new flag.
Co-authored-by: samibs <1743891+samibs@users.noreply.github.com>
This commit adds `SECURITY_ASSESSMENT_LOCAL.md`, which provides a detailed audit of Moltbot for users who want to run the assistant entirely locally using Ollama.
The assessment identifies potential data leakage points (update checks, cloud-capable tools), audits model fallback logic to prevent accidental cloud usage, verifies Gateway binding and authentication, and provides a hardened configuration template.
Co-authored-by: samibs <1743891+samibs@users.noreply.github.com>
