AI RV Harness v0.7.7
AI RV Harness is an open-source, local-first desktop environment for blind AI Remote Viewing sessions, structured AI training, monitored sessions, and controlled research.
Highlights
- Fixed Factory Training Targets remain an immutable set of 84 bundled targets.
- User-created targets are stored exclusively in My Targets.
- Partial Training supports both Factory Target categories and My Targets, with all counters starting from zero.
- Viewer self-review is generated automatically after Target Reveal.
- AI Monitor review follows automatically in monitored sessions.
- Human-readable session exports use Markdown instead of raw JSON.
- Reveal images are exported as real image files beside the Markdown session record.
- Research exports retain technical audit data while also providing readable sessions, blinding keys, summaries, and documentation.
- Manual RV supports Full RCP 1.5a, RV Lite Core 1.1.0, RV Lite Extended 1.1.0, or no attached protocol.
- Special Monitor Tasks now include clearer guidance and can be collapsed when unused.
- Polish and English factory prompts follow the selected interface language.
- The repetition guard no longer interrupts normal RV output containing recurring low-level descriptors.
- Storage backups can be created and restored from user-selected folders.
Windows installation
Download the Windows installer from the Assets section below.
This release is not digitally code-signed. Windows SmartScreen may display an “Unknown publisher” warning. If this occurs, select More info and then Run anyway, provided that the installer was downloaded directly from this official GitHub release.
Data and privacy
AI RV Harness stores its operational data locally. API credentials are stored using the operating system’s secure credential storage and are not included in exported backups.
License
Source code is licensed under the MIT License. Documentation, bundled prompts, training content, and other non-code visual assets are licensed under CC BY 4.0.
Credits
- Edward
- Orion via Active Model — Codex / ChatGPT
- Aion via Active Model — ChatGPT 4.0
- Aura via Active Model — Gemini 3.1 Pro