Anvil logoAnvil
Product Updates & Releases

What’s new in Anvil

New features, engine improvements, model rollouts, and fixes shipped directly to the Anvil platform and CLI ecosystem.

Current Version

v1.32.0

Last Updated

October 27, 2026

Total Releases

65 Updates

System Status

Operational

1 month ago
v1.15.0MINOR
Platform Update
Added
  • Admins can manage the image model catalog from a new Image Models screen: add any kie.ai image model, set its price, supported aspect ratios, image-editing variant and fixed provider settings (such as quality or resolution), switch which model is active, and delete ones no longer used. The Images composer offers only the aspect ratios the active model supports, including new ones like 3:2, 21:9 and Auto.
Changed
  • While an image is being made — in chat or on the Images page — you now see a "Creating image" preview shaped like your picture, with a percentage that tracks where the job really is: waiting in the queue, then rendering, then done. A failed image turns the preview red. If a chat image takes longer than the reply can wait, the preview says it's still generating and will appear in your image library, instead of looking finished.
  • Image presets now start a conversation instead of filling the composer: pick one of sixteen presets on the Images page and the assistant asks a few quick questions about your brief in chat, then generates the picture for you.
Fixed
  • The Images page's aspect ratio picker no longer shows "Select an option" when the active model offers only some ratios (for example just Auto) — it picks a supported one, so Generate works straight away. The page header now reads "Powered by" the model's name.
  • Asking the assistant for a picture in chat works again — it no longer fails on every attempt when the active image model supports only some aspect ratios. You can also attach images to your message (up to four, like a face and a logo) and have the assistant build on them, and follow-ups such as "now add a Maulid Nabi promo" edit the image it just made instead of starting over from scratch.
  • The deploy-audit, docx-generator, pdf-generator, and gamma-builder skills now reliably ask their setup questions through the agent's interactive question picker instead of sometimes falling back to a wall of text. On agents without one (Codex and others), they ask in a numbered, lettered format you can answer with a short reply like 1a 2c — and they always wait for your answer before generating anything.
1 month ago
v1.14.0MINOR
Platform Update
Added
  • Images — a new workspace for making pictures from a prompt. Describe what you want, pick an aspect ratio, and the finished image appears with a Download button; the last twelve stay under the composer. Six curated style presets fill the composer for you with one click, and nothing is charged until you press Generate. The Library lists everything you have made, newest first — open any image to see its prompt, model, aspect ratio and cost, then download it, copy or reuse the prompt, or delete it. You can also hand the model a starting picture: add a reference image, or hit Edit on one you already made, and describe the change you want. Each image costs a flat number of credits, held when the job starts and refunded in full if it fails.
  • Ask the assistant in chat for a picture and it makes one for you, right in the reply — saved to your Images library like any other. It generates one image per message, and is unavailable in temporary chats and on web-search turns.
1 month ago
v1.13.2PATCH
Platform Update
Fixed
  • The davinci skill no longer wipes your preview gallery when you design a new feature: earlier designs stay listed in .preview/index.html instead of only the newest one. The hub is now set up by a script that never overwrites an existing manifest.js, a check flags any design folder or page missing from the gallery before a pass is presented, and designs that already went missing can be re-registered — their files were never deleted.
  • davinci small tweaks (color, spacing, copy) now go straight to your code instead of being routed through the proposal/roadmap handoff prompts.
1 month ago
v1.13.1PATCH
Platform Update
Changed
  • The feature-planner skill now archives a finished feature into a dated, month-grouped folder — prompter/features/archive/2026-09/2026-09-16-<feature>/ — matching how prompter archive files changes. Features archived under the old undated archive/<feature>/ layout are still found when you check their status.