devlog / 2026-W40

Week 40, 2026

2026-09-28 to 2026-10-04

Merged pull requests

Project	Kenn	Community	Total	Lines added	Lines removed
AgentsView	65	33	98	+60,680	−15,232
Kenn Forge	79	2	81	+117,684	−77,518
docbank	59	2	61	+41,183	−15,822
msgvault	24	33	57	+104,253	−12,491
roborev	34	11	45	+47,325	−7,458
kata	14	22	36	+68,636	−5,264
BenchDB	11	1	12	+3,152	−293
Fotobank	4	0	4	+701	−324
kwt	2	0	2	+90	−45
Ghosthub	1	0	1	+106	−13
Total	293	104	397	+443,810	−134,460

Import sessions from Junie CLI and StepCode, SQLite-backed OpenClaw sessions, and Antigravity plaintext transcripts.
Session sync no longer loses files when sources share an ID, and Claude sub-agent continuations now stay attached to the right session.
ChatGPT re-imports now restore truncated text, fill empty tool results, append new messages, and move replaced sessions to trash.
The dashboard now shows commit and pull-request totals, activity by role, message frequency, and more sessions as you scroll.
Usage reports now separate uncached input from total input and price GPT-6 models and Codex cache writes correctly.

Connect Forge to Bitbucket Cloud and Bitbucket Data Center repositories.
ACP chats now run across execution hosts and survive Forge restarts. Chats also show tool activity, screenshots, pasted images, and image output.
Track multiple pull requests, issues, and Kata tasks in one workspace. Forge remembers workspace tabs across views and browsers.
Forge stays responsive on slow connections and usable while pull requests merge or background sync runs.
Review a pull request's images and diagrams in one expanded view, and see external context and actions beside the pull request.

Export sealed snapshots as verified load-file packages, or export selected original files through the CLI and MCP.
Import camera groups, search indexed photo metadata, and use smaller photo previews.
Create, review, download, inspect, and recover frozen search reports for selected document versions.
Imports, large directory listings, searches, storage reports, and tag pages now do less repeated work.
Backups and restores now support objects and recovery files larger than 4 GiB, and uploads no longer time out during storage work.

Sync Microsoft 365 mail through Microsoft Graph, connect multiple CardDAV accounts, and import Plaud meetings directly.
Navigate the web app with a grouped sidebar and global search. Everything and Files now share one toolbar, and you can update saved views in place.
Export original emails and full threads, manage delegated Gmail and IMAP drafts, and create native Beeper chat drafts.
Review identity matches and contact details with optional scoring. Meeting attendees now link to people in your directory.
Apple WhatsApp imports now process only new messages. Gmail sync also recovers from quota limits, and PST imports preserve reply threads.

Releases
Tag	Date
v0.21.0	2026-10-02

Use roborev doctor to diagnose setup problems.
Fix and refine can now include an optional planning phase.
The web UI now has sortable review tables, analytics by agent and model, and a light theme.
Remote daemons now support shared-key authentication and signed access to review history.
CI reviews now survive branch renames, use current fork heads, recover failed runs, and retry status delivery.

Releases
Tag	Date
v0.71.0	2026-10-03
v0.70.0	2026-09-30

Kata can now mirror Notion task databases and Plane projects as native issues, then keep issue status in sync across Kata, Notion, GitHub, and Plane.
GitHub imports now show progress, filter imported history, and find parent links incrementally instead of rescanning the whole repository.
Coding agent setup now supports more native integrations, avoids duplicate Codex hooks, and lets Muse work without hooks. OpenClaw inbox sharing is now opt-in.
You can now run kata doctor to diagnose local setup problems. Search also reports missing or rejected embedding credentials.
Federated setups can now sign requests and restrict incoming connections. Local metadata writes also stay responsive when hub lease refresh stalls.

BenchDB now saves completed benchmark reports, making repeat reads less expensive.
You can now deploy BenchDB in a standard container using published release images.
Configurable timing and memory tolerances can hide benchmark changes too small to matter.
Saved run reports now honor the selected baseline, and individual results can opt out of statistical history.
Benchmark charts are easier to find and share, and chart cards now load the complete result history.

Telemetry now reports a development version for local and nightly builds instead of treating them as release builds.