AutoFigure-Edit v1.1

AutoFigure-Edit v1.1

This release is published as tag v1.1.

AutoFigure-Edit v1.1 focuses on two practical workflows that were still awkward in earlier public builds: starting from a user-supplied stage-1 academic figure, and running the pipeline cleanly with official OpenAI models or OpenAI-compatible gateways.

Highlights

1. Stage-1 figure import workflow

• Added a dedicated import-mode path in the web UI for users who already have the stage-1 academic raster figure.
• Added and documented the CLI path for skipping step 1 image generation and continuing directly from SAM + SVG reconstruction.
• Imported figures are normalized into the regular pipeline output structure so later SAM, placeholder, and SVG steps behave the same as a full run.

Why this matters:

• You can now iterate on an existing figure instead of regenerating from method text every time.
• The product is more practical for users who already have a draft figure from another model, another provider, or a previous run.

2. Official OpenAI model support

• Step 1 can now use the official OpenAI Images API with gpt-image-2.
• The OpenAI Responses route is documented and exposed for text plus multimodal SVG reconstruction.
• openai_response now defaults to gpt-5.5 for the SVG / reasoning path.

Why this matters:

• The OpenAI path is now a first-class documented workflow instead of an implicit or partial configuration.
• Users can mix OpenAI image generation with OpenAI multimodal reconstruction more predictably.

3. OpenAI-compatible routing and custom provider support

• Added custom as the primary OpenAI-compatible provider name in the CLI and web app.
• Kept bianxie as a backward-compatible alias so older commands do not break.
• Fixed the default openai_response main route so step 1 can inherit the same compatible base_url and api_key instead of incorrectly falling back to the official OpenAI host.

Why this matters:

• A single compatible gateway can now drive the whole workflow more reliably.
• Users no longer need to fight inconsistent routing between the image stage and the SVG reconstruction stage.

4. Web configuration, bilingual UI, and onboarding updates

• Added bilingual Chinese / English switching across the main page, import page, canvas page, and guide page.
• Added an in-product configuration guide that explains workflow choices, field meanings, recommended presets, and SAM backend setup.
• Kept the release aligned with the newer DeepScientist-branded web surface.

Why this matters:

• Setup is easier to understand for first-time users.
• The import flow and provider flow are now visible in the product instead of being hidden in CLI-only assumptions.

Included Areas of Change

• Pipeline and provider routing: autofigure2.py
• Web request handling: server.py
• Web UI and bilingual configuration flow: web/index.html, web/import.html, web/guide.html, web/app.js
• Documentation updates: README.md, README_ZH.md

Release Notes

• Git tag: v1.1
• GitHub release: AutoFigure-Edit v1.1
• Default OpenAI image model: gpt-image-2
• Default OpenAI Responses SVG model: gpt-5.5
• No standalone binary assets are attached to this release; GitHub provides the source tarball and zipball for tag v1.1.