What's new for you
Dreaming and memory-wiki: ChatGPT import ingestion with Imported Insights and Memory Palace diary subtabs; Control UI / webchat: structured bubbles for media/reply/voice, `[embed ...]` rich output with gated URLs; `video_generate`: URL-only asset delivery, `providerOptions`, reference audio, adaptive aspect, higher image-input cap; Feishu document-comment threads with reactions; Teams reactions with Graph pagination and delegated OAuth for sends; plugin manifests gain activation/setup descriptors; Ollama caches `/api/show` context during discovery; plus Codex OAuth scope fix, transcription DNS, Talk Mode mic permission, WhatsApp default account and reactions, and many agent/QA fixes.
- Dreaming — ChatGPT import, Imported Insights & Memory Palace subtabs in diary UI
- Control UI — structured assistant bubbles, `[embed ...]` tag, gated external embed URLs
- video_generate — URL delivery, typed providerOptions, reference audio, adaptive aspect
- Feishu — document comment sessions with reactions and typing feedback
- Teams — reactions, listing, delegated OAuth for reaction sends
- Plugins — manifests declare activation & setup descriptors for onboarding flows
- Ollama — cache model discovery metadata; invalidate on digest changes
Release highlights
| Feature | Details |
|---|---|
| Dreaming: ChatGPT import & Memory Palace | Ingest ChatGPT exports into memory-wiki; new Imported Insights and Memory Palace diary subtabs so Dreaming can inspect source chats, compiled wiki pages, and full source pages in the UI. |
| Control UI / webchat: bubbles & embed | Assistant media, reply, and voice directives render as structured chat bubbles; [embed ...] rich output tag; external embed URLs gated by config. |
| video_generate: URL delivery & options | URL-only generated asset delivery, typed providerOptions, reference audio inputs, per-asset role hints, adaptive aspect ratio, and a higher image-input cap for richer provider modes. |
| Feishu document comments | Richer context parsing, comment reactions, and typing feedback so document-thread conversations behave more like chat. |
| Teams reactions & delegated OAuth | Reaction support, reaction listing, Graph pagination, and delegated OAuth for sending reactions while preserving application-auth read paths. |
| Plugin activation descriptors | Plugin manifests can declare activation and setup descriptors so setup flows describe auth, pairing, and configuration without core hardcoding. |
| Ollama discovery cache | Caches /api/show context-window and capability metadata during model discovery; retries after empty responses; invalidates on digest changes. |
Feature deep dives
Learn more about the changes that matter most. Bug fixes and minor improvements are listed on GitHub.
Dreaming: ChatGPT import & Memory Palace
Import ChatGPT chats into memory-wiki; Imported Insights and Memory Palace subtabs in the Dreaming diary.
Read more →Webchat bubbles & embed tag
Structured bubbles for media/reply/voice; [embed ...] rich output with config-gated external URLs.
Read more →video_generate delivery & options
URL-only assets, providerOptions, reference audio, adaptive aspect, higher image-input caps.
Read more →Feishu document comment threads
Reactions and typing feedback on document threads for chat-like collaboration.
Read more →Teams reactions & OAuth
Reactions, listing, Graph pagination, delegated OAuth for reaction sends.
Read more →Plugin activation & setup descriptors
Manifests describe activation, auth, pairing, and setup without core special cases.
Read more →Ollama model discovery cache
Cache /api/show metadata across picker refreshes; invalidate when digest changes.
Read more →Full release notes
The complete changelog with all changes, fixes, and technical details is on the official GitHub release page.
Open v2026.4.11 on GitHub →