open-design open-design-v0.20.1

open-design-v0.20.1

Open Design 0.20.1 — Meet Go Plan: Unlimited Access to Three Powerful Models

Added 5
  • Introduce Go Plan with unlimited access to DeepSeek V4, DeepSeek Pro, and GLM 5.2
  • Add Go column to pricing page and launch banner on homepage with in-app upgrade entries
  • Add handoff to OpenCode community plugin to continue work in preferred agent
  • Add image generation showcase on landing page
  • Add DeepSeek Harness GUI guide with steps 2–5 walkthrough
Changed 4
  • Make plan names, order and copy consistent between app and website
  • Standardize product name to OpenDesign across app and site
  • Optimize conversation list to no longer materialize full event logs for rendering
  • Anchor run timing to moment model starts responding instead of including queueing and startup
Fixed 11
  • Allow models with unlimited plan access to remain available even when wallet balance is zero
  • Display specific error codes and reasons for image generation failures instead of generic message
  • Display specific error codes for deploy failures instead of opaque 400 error
  • Preserve About panel spacing and prevent project frame from flashing twice on open
  • Remove cleared skills from @ mention picker instead of leaving stale entries
  • Coalesce replay events when reattaching to running conversation

From open-design

🌟 Codename: Meet Go Plan: Unlimited Access to Three Powerful Models

🟢 28 PRs · 10 contributors · 2 days — Create without worrying about usage limits. Go Plan gives you unlimited access to DeepSeek V4, DeepSeek Pro, and GLM 5.2—so you can explore, create, and iterate as much as you want.

🔥 Highlights

  • 🟢 Go is here — a plan you can actually start on. The new entry tier ships end to end: a Go column on the pricing page, a launch banner on the homepage, and in-app upgrade entries that take you straight to Pricing instead of a dead-end dialog. Plan names, order and copy now read the same in the app and on the site. (#7121, #7122, #7167)

  • 🔋 A zero balance no longer locks you out of the models you pay for. If your plan includes unlimited use of a model, the picker keeps offering it even when your wallet reads 0 — the balance and the entitlement are two different things, and only one of them should gate the model list. (#7187)

  • 🖼️ When an image doesn't come back, you find out why. Generation failures used to collapse into one generic message. Now the real reason comes through — a safety refusal, a provider error, a bad prompt — so you know whether to retry, rephrase, or switch models. (#7118)

  • 💬 The first thing you ask for on Home streams all the way through. Starting a run from Home could leave the output frozen partway while the work continued in the background. The first run now stays live from the first token to the last. (#7071)

  • ⏱️ "How long did that take" finally means what it says. Run timing is now anchored to the moment the model actually starts responding, so queueing and startup no longer inflate generation time in your history. (#7155)

  • 🧩 Handoff to OpenCode. A new community plugin hands the current piece of work over to OpenCode, so a task started in Open Design can continue in the agent you prefer. (#6948) Thanks @ross-sec.

  • 🏷️ One name, everywhere: OpenDesign. The product name, navigation and CTAs now read consistently across the app and the site, instead of drifting between two spellings. (#6983, #6998) Thanks @joeylee12629-star.

  • ⚡ Opening your conversation list stops doing unnecessary work. The list no longer materializes full event logs just to draw itself, which shows up directly as a faster panel on projects with long histories. (#6978) Thanks @jiulongche.

🩹 Fixed
  • The About panel's update row keeps its spacing, and the project frame no longer flashes twice when a project opens. (#7174)
  • Deploy failures return specific error codes instead of one opaque 400, so a failed deploy says what went wrong. (#5897)
  • Removing a skill clears it from the @ mention picker instead of leaving a stale entry behind. (#6920) Thanks @T-sanjay-ram.
  • Reattaching to a running conversation coalesces replay events instead of replaying them one by one. (#6926) Thanks @roian6.
  • Agent runtimes start on Windows machines whose PATHEXT is malformed. (#6937) Thanks @github-lover-9999.
  • od media wait and od media generate flush their output before exiting, so piped and scripted usage no longer loses the last lines. (#6800) Thanks @lorenzozanee.
  • The artifact upsell only appears for Cloud models, not for models you already brought yourself. (#6995)
  • Preview font stylesheets load without blocking the preview itself. (#7134)
  • The DeepSeek Harness bootstrap version is derived from the installer bytes, so the reported version matches what actually installed. (#7107)
🌐 Website
  • An image generation showcase on the landing page. (#6904)
  • A DeepSeek Harness GUI guide that answers the search intent directly, with the steps 2–5 walkthrough split out. (#7073) Thanks @joeylee12629-star.
  • DeepSeek curated plugin detail routes work again. (#6976) Thanks @roian6.
🙏 Thanks to everyone who shipped 0.20.1

@alchemistklk · @github-lover-9999 · @jiulongche · @joeylee12629-star · @lefarcen · @lorenzozanee · @roian6 · @ross-sec · @Siri-Ray · @T-sanjay-ram

View original

Upgraded? How did it go?

Discussion