Skip to content

Image2Tools v0.2.0 (unsigned preview)

Pre-release
Pre-release

Choose a tag to compare

@Bliveren Bliveren released this 11 Jun 09:31

Image2Tools v0.2.0 Release Notes

Status: unsigned preview prerelease. macOS and Windows artifacts are published for evaluation. macOS builds are ad-hoc signed and not Apple-notarized, so Gatekeeper may warn on first launch (right-click > Open, or remove the quarantine attribute). Developer ID signed and notarized distribution is tracked for a future release.

Highlights

  • Multi-model workspace for GPT Image 2, Nano Banana 3, and General launch flows.
  • Refined left sidebar with separate model configuration, launch model, parameter configuration, draft, notice, and update areas.
  • Model configuration now runs automatic connection checks on startup and after saved API config changes. The UI shows connection status beside the model configuration title and surfaces friendly failure guidance.
  • Model discovery can infer provider/model availability from the configured API and enables launch families based on discovered models rather than only the selected provider label.
  • OpenAI provider path keeps GPT Image 2 generation, editing, exact-mask inpainting, streaming partial previews, downloads, and history.
  • GPT Image 2 multi-output generation now uses non-stream requests when count is greater than one, backfills providers that return fewer images than requested, and shows selectable final-result thumbnails in the canvas.
  • Gemini provider path adds the app's Nano Banana 3 launch target for discovered Gemini image models such as gemini-3.1-flash-image and gemini-3-pro-image, with generateContent image generation, reference-image editing, guided-region editing, and Gemini-specific controls.
  • Launch model buttons now include concrete discovered-model choices when multiple compatible models are available under the same launch family.
  • General launch mode is a minimal fallback for discovered non-focused image models. Gemini General supports prompt and reference images; OpenAI and Custom General use a prompt-only OpenAI-compatible /images/generations contract.
  • History entries now retain provider/model context so reused prompts restore the matching launch model and parameters.
  • Clearing all recent-task history now requires explicit confirmation.
  • Update checks now default to the raw GitHub manifest URL to avoid the previous GitHub Pages 404. The sidebar shows an automatic up-to-date status, saves the current draft before update launch, and starts the Windows NSIS installer in silent mode with app restart after installation.

Verification Gates

Before publishing v0.2.0, run:

pnpm build
pnpm verify:mock-api
pnpm verify:mock-gemini-api
pnpm verify:mock-model-discovery

OpenAI real API acceptance is cost-gated through:

IMAGE2TOOLS_API_KEY=sk-... IMAGE2TOOLS_REAL_API_ACCEPT_COST=1 pnpm verify:real-api

Gemini / Nano Banana 3 real API acceptance is cost-gated through:

IMAGE2TOOLS_GEMINI_API_KEY=... IMAGE2TOOLS_REAL_GEMINI_API_ACCEPT_COST=1 pnpm verify:real-gemini-api

The verifier covers discovery, generation, reference editing, and guided-region editing. Do not mark Gemini real acceptance complete until the verifier and app-level history/download checks pass with a real Gemini key.

Release package verification remains platform-specific:

pnpm verify:release:mac
pnpm verify:release:windows
pnpm verify:release:linux

Re-run the secret scan in SECURITY.md after every release-note or packaging metadata update.

The app package metadata is 0.2.0. For this unsigned preview, docs/updates/latest.json
publishes the darwin (arm64 dmg) and win32 (x64 installer) assets with verified URL,
sha256, and sizeBytes metadata generated from the exact uploaded artifacts. When
Developer ID signed and notarized macOS artifacts become available, regenerate the
manifest from the signed build and supersede this preview.

What's Changed

Full Changelog: v0.1.0...v0.2.0