Cleverly v1.0.250

Installation

Choose the artifact for your platform:

Platform	File	Notes
Windows	Cleverly-Setup-*.exe	NSIS installer — installs app + adds to Start Menu
Linux	*.deb	Debian/Ubuntu — auto-pulls system deps (Python, ffmpeg, GTK)
Linux	*.AppImage	Universal — no install, just chmod +x and run
macOS	*.dmg / *.zip	Intel and Apple Silicon builds; unsigned, so first launch may require approval in Privacy & Security

macOS builds are not code-signed or notarized. Gatekeeper may block first launch; users may need to approve the app in System Settings → Privacy & Security. For distribution to general users, signing and notarization are recommended.

First Run

On first launch, Cleverly opens the Settings window automatically for your Google Gemini API key.
Get a free key from Google AI Studio. You can also edit .env directly.

Optional: Local Whisper (offline speech)

Install Python 3.10+ and ffmpeg, then run:

./setup.sh --no-run

Checksums

See SHA256SUMS.txt attached to this release.

What's Changed

Summary

This release includes 38 commits since v1.0.212.

Features

• Add runtime endpoint safety invariant and diagnostics for hosted AI routing (e2bb3cc)

Fixes

• Fix spoken assistant regression assertions (dfa31c6)
• Fix spoken assistant regression assertions (4201482)
• Fix fast-chat regression test syntax (7af9450)
• force deterministic desktop source selection in E2E (3f94e27)
• run gated release after successful main-branch verification (820bbd1)

Other Changes

• Make typed chat regression test cross-platform (26d8a86)
• Route spoken chat through the general Cleverly assistant (a9de25f)
• Cover general assistant routing for spoken chat (534f7f8)
• Require explicit release trigger after validation (95983b5)
• Tag hosted screen-vision requests separately (c3c5203)
• Budget screen vision separately from chat traffic (a95f85a)
• Fingerprint screen-specific usage budget (5cb15f6)
• Test dedicated screen-vision request budget (adeec2c)
• Throttle continuous screen analysis to protect provider quota (7cd5143)
• Test isolated screen vision and chat budgets (e0b08df)
• Align live screen test with quota-safe frame interval (218bff9)
• Restore v175 single-request fast path for chat (94beb65)
• Verify v175 fast chat path and screen-only fallback (8bb62f2)
• Fingerprint v175 fast-chat behavior for deployment verification (9db3e79)
• Verify deployed v175 fast-chat API revision (81d0ac6)
• add explicit v175 fast-chat and screen-vision verification (cc308fb)
• align screen capture assertion with hosted request budget (2f508bf)
• trace screen AI buttons through IPC and Gemini vision (236735b)
• run screen AI UI route checks in regression suite (cf2df6e)
• add real desktop screen AI end-to-end coverage (64f8904)
• enable deterministic display selection for desktop E2E only (ed1c5d1)
• add desktop screen AI E2E command (9d94ee1)
• run real desktop screen AI UI E2E under Xvfb (0d690fd)
• gate builds and releases on real desktop screen AI E2E (123c1b5)
• pass Electron flags before app entry point (2b82f35)
• allow local CDP websocket in desktop E2E (bd74efa)
• make desktop UI E2E deterministic with a synthetic display stream (34dd911)
• route desktop screen AI E2E through hosted Gemini proxy (bbe33c1)
• always emit Electron diagnostics for failed screen AI E2E (2cfd148)
• Force screen AI requests through fixed hosted endpoints (9e87df6)
• Use Node URL constructor for hosted AI endpoint validation (ee4238a)
• Allow hosted AI HTTPS fallback enough time for cold starts (e2fd3fa)

Full Changelog: v1.0.212...v1.0.250