Ships

Make the chat PAGE the browser agent (no background worker)

Render in-page browser via reliable room socket (web_navigate), not just the extension

In-page browser: force iframe reload on every navigate (nonce)

Don't cache the /chat HTML shell (no-store)

In-page browser: React-controlled iframe src (fixes blank panel)

In-page AI browser: render the browsed site as an iframe in the chat Web panel

Web browsing: feed search/browse results back to the agent

Web browsing: add Grok function-calling tool defs + tool-call translation

Add AI web browsing: {search}/{browse}/{web_click} commands routed to user's browser via new Chrome extension (packages/browser-extension), cmd_web room toggle, Web panel tab

Add @commslink/hologram package with Phase 1 demo route

Chat: volume slider is now live — affects all in-flight audio (AI, own voice, other users)

Dashboard layout header: show 'CommsLink Chat' (purple) on /chat

Header: show 'CommsLink Chat' (purple) on /chat

Dashboard: add Evolution & Fitness section with generation history, fitness by school, and gen badges on leaderboard

CommsLink login, persistent chat history, voice

Boredom emotion + LLM dreams evolve freely. Bored farmers explore, socialize, take risks. Vision not locked to genome

Family bond system: trust decay, passing nods, context greetings (spouse/child/parent/kin), social anxiety drives relationship repair

Structured memory system: categories (relationship/life_event/skill/belief/place), compact index, selective recall, context-aware talk

Fix watch party pause: update local state immediately on custom control click

Watch party: video full-size in main area, chat moves to side panel tab

Watch party: paste YouTube URL in chat to start synced video for the room

Army chat spacing commands: tighten up/phalanx, normal spacing, spread out/loose

Formation system: 7 formations, calculator, behavior tree, API, chat commands

Speed up army chat: fast pattern matching for commands (34-159ms vs 12s Grok)

Add dispatcher routing tests (13 tests) + army chat wiring instructions for Unity

Army chat: split into fast dispatch + async WebSocket responses

Army chat: return responder + listener ID lists so client shows thinking bubbles only on responders

Army system backend: hierarchy, chat dispatcher, commands, promote, reassign

Army system: hierarchy, chat dispatcher, commands, reassign, promote, auto-assign on recruit

Add army system client instructions: hierarchy, chat routing, overview panel, formations

Add POST /api/v1/recruits/:id/chat — talk to recruits via Grok

Game world: full AI companion system with behavior tree, Grok brain, NPC engine

AI companion system: personality traits, disposition, behavior weights, Grok brain, behavior tree, recruit instructions/commands API

Add initRoom debug logging for chat connection flow

GameWorld: auto-add players to global-game chat room on game:join

Logo/title click navigates to /landing instead of /chat

FFXIVoices: dedup chat messages — prevent double TTS when multiple clients send same text

FFXIVoices API: auth, chat/TTS, WebSocket audio broadcast, ElevenLabs adapter, proximity filtering

Strip leaked [/add_memory] and other bracket-format closing tags

Strip leaked tags: xai:function_call XML, orphaned /say /text, unmatched command braces

Reduce thinking filler trigger from 3s to 2s

Thinking filler: grok-3-mini picks contextual phrase, spoken via Polly TTS

Make logo + CommsLink title clickable, navigates to /chat

Add thinking filler message if AI takes >3s to respond

Fix think leak: strip all malformed think tag variants ([/think>, mixed brackets, etc)

Mobile: terminal panel stacks above chat in top 1/3, hide resize handle

Fix: keep Claude Code responses in agent context (was filtering all [Claude *] messages)

Add + create room button to sidebar on all pages, navigates to chat with create dialog

Show room icons in sidebar on all pages (credits, profile, etc), not just chat

Remove Chat tab from sidebar — show room icons directly, cleaner layout

Fix credits page: center content, hide offline indicator on non-chat pages

Fix think-only: if AI only thinks and doesn't speak, retry with nudge to respond

Add max tokens slider to agent create/edit, default 2500, range 500-8000

Default ON: memory, think, intent coherence, memory coherence for new rooms

Fix dedup: raise threshold to 0.8, shorten window to 5min — stops blocking normal responses. Louder thinking sound

Add thinking sound (gentle pulse) and improved typing indicator while AI generates response

Voice debounce: 3s delay before AI responds to voice, waits for user to finish speaking

Fix: STT messages now trigger AI agent responses (mention detection + alone-in-room)

Fix: say/text always available even when think is OFF — agents must always be able to speak

20 agent templates, increase agent limit to 5 per room

Fix voice selector gap: remove minWidth, disable autoWidth

Fix voice: read user voice_id from DB on each transcript, not just session start

Voice: server handles everything — broadcast message + TTS directly from transcript callback, no client round-trip

Fix voice: emit directly in onTranscript callback, bypass sendMessage stale closure

Don't pause voiceCapture during audio playback — echo filter handles it, pausing kills the stream

Voice selector: show short name in toolbar, full name in dropdown, min 70px width

Fix TTS: fallback invalid voice IDs to Joanna, updated all users/agents in DB

Fix TS2345: non-null assert on data.voice

Server-side TTS for user messages, fix duplicate transcripts, remove client TTS calls

Fix duplicate text, add TTS debug logging, deduplicate transcripts

Fix: 5s flush for long speech, increase TTS text limit to 5000, mobile voice selector sizing

Real-time voice: text sent immediately, TTS generated in background, chat_audio broadcast

Skip TTS generation for voice-captured messages — user already spoke them

Real-time STT: streaming Transcribe session, transcripts arrive as you speak

STT debug logging: chunk sizes, total bytes, transcript output

Voice mic: show errors as toast, fall back to Web Speech API on failure

Fix double message: voiceCapture takes priority over voiceStream, remove duplicate path

Voice pipeline: MediaRecorder STT (no beeps), Amazon Transcribe, Comprehend sentiment, emotion-aware Polly TTS

Add Claude API, fix voice echo (pause mic during AI playback), increase restart delay

Terminal panel: creator-only downloads and input, informational view for non-creators

Terminal onboarding: download buttons in empty state, localStorage persistence, auto-open for new rooms

Move 'hear my own voice' to room settings, remove settings gear icon and panel

YouTube-style volume slider, remove stop audio button

Add tiny volume slider in chat input toolbar next to voice selector

Add voice selector dropdown in chat input toolbar next to mic button

Fix invite link: switch room immediately on connect instead of unreliable timeout

Default voice to Joanna when none selected

Add echo cancellation: filter mic input that matches recent AI speech output

Remove browser voices entirely, Polly only

Replace ElevenLabs with Amazon Polly TTS — 75x cheaper, same AWS region

Remove alarm button, fix settings panel mobile overlay with safe area

Mobile safe area: status bar padding, correct chat height, viewport-fit cover

Make website mobile-responsive: compact header, responsive activity bar, fullscreen panels on mobile, hide format toolbar

Remove web browser, watch party, watchlist, screen share, scheduled jobs, daily stats, about page

Add close button to settings panel

Remove cost calculator display, spending bar, credits page, and payment routes from UI

Default off: intent/memory coherence, think, effort, tokens, audit, extended thinking

Add toggleable Intent Coherence and Memory Coherence settings in room settings

Remove entire hologram system

Redesign credits page with usage graph, time range filter, service breakdown