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

2 months ago
v1.9.1PATCH
Platform Update
Changed
  • Only the main Chat assistant remembers things about you now. The specialist assistants still use what is already remembered, but no longer save new facts of their own.
Fixed
  • Chat can save new memories again. Saving had stopped working for the main assistant, so facts you shared were never stored, even though memories from before were still used.
2 months ago
v1.9.0MINOR
Platform Update
Added
  • Admins can set maximum output tokens in Prompt CMS Routing Rules to allow longer answers.
  • Attach, paste, or drop up to four images into Chat, preview them before sending, and include them in shared conversations.
  • Admins can enable or disable image support for each Chat model from Admin → Models.
  • Choose Search for quick web answers or Deep research for a more thorough response with sources.
  • Review Chat prompt enhancements before replacing your draft, request targeted revisions, and see the credit cost before regenerating.
Changed
  • Chat now uses the standard assistant experience without showing AI persona controls.
Fixed
  • Long questions now collapse with Show more and Show less controls in Chat and shared conversations.
  • New conversations accept text and image-only messages without incorrectly reporting that the message or model is missing.
  • Chat now receives the current date on every turn and guidance to verify time-sensitive search results.
2 months ago
v1.8.0MINOR
Platform Update
Added
  • You can now start a temporary chat for a throwaway conversation. It deletes itself 24 hours after its first message, shows its remaining time in the chat and sidebar, cannot be pinned or shared, and never saves new memories about you.
  • Chat replies now show the model that answered, with its provider mark, so it is easier to see which model produced a response.
Changed
  • The rich-text editor now matches the typography and formatting of its rendered preview more closely, including headings, lists, quotes, code, links, and tables.
2 months ago
v1.7.0MINOR
Platform Update
Added
  • Before saving something new to memory, the assistant now checks what it already knows about you. A fact it already has on file is recognized rather than stored a second time, and a fact that revises an older one is merged into it instead of leaving both versions on record — so your memory list stops filling up with near-duplicates and contradictions. This costs you no credits, and if the check ever fails the fact is simply saved as before, never dropped.
  • The Memory section in settings now shows when a memory was last confirmed alongside when it was last updated, so you can see which facts the assistant keeps finding still true and which have not come up since they were written.
Changed
  • When your memory is full, the assistant now names the specific memory that has gone longest without coming up as the one to consider forgetting, instead of a generic "memory is full" message. You still decide whether it goes.