Skip to content

Releases: eliautobot/my-virtual-world

My Virtual World v1.0.23

Choose a tag to compare

@eliautobot eliautobot released this 08 Jul 02:27

Highlights

  • Expanded Live Agent Mode into a continuous planner with reflection, current goals, multi-step plans, next steps, safe requests, and memory updates.
  • Improved Live Agent chat readability with structured planner cards and collapsible raw planner JSON.
  • Added a Sessions panel in the chat window for selecting, creating, and deleting agent sessions directly.
  • Added unified session management across supported agent frameworks, including OpenClaw, Hermes, Codex, and Claude Code session records where configured.
  • Improved chat history loading so each selected session shows its own conversation or thread history.
  • Preserved session metadata in chat bubbles so normal chats, Live Agent Mode, and provider-backed sessions remain clearly labeled.
  • Improved realtime Live Agent action visibility, completion state, route handling, and user-preemption behavior.
  • Added embodied seating validation so arrival alone is not counted as successful object use.
  • Added building placement safeguards to prevent structures from overlapping streets, curbs, or sidewalks.
  • Added broader verification coverage for Live Agent autonomy, session handling, realtime behavior, construction routing, provider session handling, and placement rules.

Live Agent Mode

  • Live agents now maintain a more continuous autonomy frame instead of a one-shot action picker.
  • Planner turns preserve reflection, goals, planned steps, next actions, target criteria, success criteria, and memory lessons.
  • Live Mode now carries active plans forward between ticks and records useful reflections without exposing unsafe execution paths.
  • User messages can temporarily preempt Live Mode work so direct operator interaction stays higher priority than autonomous activity.
  • Live-world ownership checks reduce conflicts when an agent is already claimed by another active world.
  • Seating/object-use validation now requires embodied completion evidence before marking a target as tested.

Chat And Sessions

  • The chat window now includes an integrated Sessions drawer.
  • Users can switch between available sessions without leaving the world UI.
  • Users can create fresh sessions or delete old sessions from the same chat surface.
  • OpenClaw sessions, Hermes sessions, Codex threads, and Claude Code session records are normalized into one chat-session UI where available.
  • Live Agent Mode appears as a readable session with structured planner replies.
  • Provider-backed histories keep their own session/thread context instead of being flattened into one generic chat log.

World Editing

  • Building saves now enforce placement rules against street surfaces.
  • Building footprints may be placed next to streets but cannot overlap roadway, curb/gutter, or sidewalk areas.
  • Placement-rule guidance is exposed through the API and documented for agent/tool users.

Verification

  • Public smoke suite
  • Realtime smoke suite
  • Live Agent autonomy verifier
  • Chat-session verifier
  • Building-placement rules verifier
  • Live Agent resident-profile verifier
  • Live action construction-routing verifier
  • Live Mode layer-separation verifier

Known Limitation

  • Resetting a normal chat session does not reset Live Agent runtime state yet. A dedicated Live Agent state reset should be added separately.

v1.0.22 - Agent profiles and global model provider settings

Choose a tag to compare

@eliautobot eliautobot released this 07 Jul 02:46

Highlights

  • Adds the My Virtual World Agent Character Builder workflow for editing agent identity, assignments, status, needs, appearance, workspace docs, resident profiles, and skills.
  • Introduces MVW Resident Profiles as the Live Agent Mode memory/personality layer separate from native OpenClaw/Hermes/Codex identity files.
  • Adds a themed Models & Providers window for OpenClaw, Hermes, Codex CLI, and Claude Code setup.
  • Brings OpenClaw model switching into MVW, including per-agent model assignment, /session-info?agent=... model reporting, and gateway restart requests after model/config changes.
  • Adds MVW Global Auth for OpenClaw static API-key/static-token profiles, with main as the source, Sync Static Profiles, per-agent auth store drift reporting, Reset Static Overrides, and OAuth preservation.
  • Improves Models & Providers UX with clear explanations, synced-profile summaries, MVW confirmation dialogs, and scroll-preserving refresh after save/apply/reset/delete actions.
  • Improves mobile and performance behavior: searchable mobile agent picker, gzip static serving, smaller mobile startup chunk radius, and read-only chunk hydration.
  • Improves OpenClaw chat failure handling so provider/model/auth failures surface cleanly instead of leaving the chat stuck thinking.

Verification

  • GitHub Smoke Test passed for commit 192f95d.
  • Local npm run verify:smoke passed.
  • Local npm run verify:realtime passed.
  • Live Docker app rebuilt/restarted on 8590 and /healthz returned OK.
  • Live Models & Providers page served MVW Global Auth, mvw-confirm-overlay, and scroll-preserving refreshNativeSoon code.
  • Browser/CDP checks verified the MVW confirmation dialog opens/cancels and action refresh preserves scroll position.
  • OpenClaw auth sync behavior was tested against isolated fake SQLite auth stores before live deployment.

Upgrade Notes

  • Docker installs now mount the OpenClaw home read-write so MVW can persist model/provider/profile changes.
  • Static API-key/static-token OpenClaw auth profiles are treated as MVW-managed global profiles from main.
  • OAuth profiles are intentionally not copied across agent stores; they remain interactive/local or inherited from native OpenClaw behavior.
  • Claude Code setup fields are present, but live Claude Code-backed agent runtime support is still pending provider runtime integration.

v1.0.21 - Server runtime parity and server-owned ping-pong

Choose a tag to compare

@eliautobot eliautobot released this 03 Jul 21:02
2135470

Highlights

  • Ships the server-authoritative Live Agent Mode runtime stack with the Colyseus realtime sidecar.
  • Adds shared runtime snapshots, route leases, observer buffering, visual state mirroring, and world runtime authority.
  • Brings server-owned agents closer to the 8590/8595 behavior baseline with collision routing, route smoothing, door repair, queue coordinate fixes, route watchdogs, social conversations, desk fallback, object orientation fixes, armchair/dismount handling, and runtime cleanup.
  • Moves ping-pong fully to server-pingpong-runtime, including server-owned pairing, ball/score state, player leases, cooldowns, stale cleanup, browser read-only rendering, and clean result/dismount transitions.
  • Makes the Docker install product-ready by starting both virtual-world and virtual-world-realtime, documenting private port 8591, and removing PR-specific cache keys.

Verification

  • Product smoke suite passed on GitHub.
  • Fresh PR #59 GitHub checkout was tested locally before merge with smoke, realtime, collision, door, route, queue, watchdog, social, Python compile, Docker Compose, Docker build, and live 8587/8588 review checks.

Upgrade Notes

  • Docker Compose now starts a realtime sidecar service named virtual-world-realtime by default.
  • Keep the web port 8590 and realtime sidecar port 8591 private to your machine, LAN, VPN, Tailnet, or authenticated reverse proxy.

v1.0.20 - Roadway drawing improvements

Choose a tag to compare

@eliautobot eliautobot released this 18 Jun 00:10

Roadway editor and rendering improvements for the 8590 Virtual World build.

Highlights:

  • Improved roadway line drawing with centerline snap, auto-alignment, and faded live roadway previews.
  • Added roadway edit handles for endpoints and midpoints, including segment extension from endpoint handles.
  • Smoothed turn, T-intersection, and partial-intersection curb/gutter continuity.
  • Cleaned up rounded sidewalk/curb corner rendering so connections sit flush and avoid overlapping perpendicular pieces.
  • Updated smoke verification for the new main3d cache key.

Verified:

  • node --check src/client/js/main3d.js
  • git diff --check
  • npm test
  • live 8590 visual checks during roadway iteration

v1.0.19 - Codex context display

Choose a tag to compare

@eliautobot eliautobot released this 17 Jun 04:06

v1.0.19 - Codex context display

This release fixes the Codex model/context display in chat windows.

Changes

  • Codex model labels now use the same blue styling in cloned/secondary chat windows as the primary chat window.
  • Codex context usage now updates from Codex native token_count session logs instead of staying at a dash.
  • /session-info for Codex agents now returns contextUsed, contextWindow, codexContextWindow, and token usage details when available.
  • Codex SSE terminal events now include context usage so the model bar can update immediately after streamed turns.
  • Added a portable fallback to use Codex's own model_context_window when OpenClaw model config does not define a context window.
  • Cache-busted chat.js and style.css for the updated UI behavior.

Verified

  • npm test
  • python3 -m py_compile src/server/server.py
  • node --check src/client/js/chat.js
  • git diff --check
  • Live Codex SSE smoke tests on 8590 and 8586 before release

v1.0.18 - Codex streaming

Choose a tag to compare

@eliautobot eliautobot released this 17 Jun 02:10

Changes

  • Add native Codex app-server run streaming endpoints for live tool-call and assistant delta events.
  • Update chat UI to use EventSource for Codex agents, render live tool cards, and support streamed stop/interrupt.
  • Persist streamed Codex results back into Codex chat history.
  • Add smoke coverage for Codex stream server/client/provider wiring.

Verification

  • npm test
  • Python compile check
  • node --check src/client/js/chat.js
  • Live 8590 SSE smoke emitted run.started, tool.started, message.delta, and run.completed.

My Virtual World v1.0.17

Choose a tag to compare

@eliautobot eliautobot released this 16 Jun 19:53

What's Changed

  • Added native Hermes API streaming as the primary chat path through POST /api/hermes/runs and GET /api/hermes/runs/{run_id}/events.
  • Proxied Hermes native run approvals and stop requests while keeping API keys server-side.
  • Added product-neutral Hermes API settings, Docker environment defaults, and documentation for /v1/runs, SSE events, approvals, and stop.
  • Kept /api/hermes/chat and /api/hermes/live as CLI compatibility fallbacks when the native API server is unavailable.
  • Bumped package version to 1.0.17.

Verification

  • git diff --check
  • python3 compile for src/server/server.py and src/server/providers/hermes.py
  • node --check src/client/js/chat.js src/client/js/settings.js
  • npm test
  • Live source install: http://127.0.0.1:8590/healthz
  • Live source install: /api/hermes/test reported native run submission, SSE, approval, and stop support
  • Live source SSE probe completed with streamed message.delta events
  • Separate GitHub install on port 8586 rebuilt from commit 63779fc
  • Separate GitHub install SSE probe completed: run_12980ff8fbcf439da70b350fe7bbdb6c

My Virtual World v1.0.16

Choose a tag to compare

@eliautobot eliautobot released this 16 Jun 15:55

What's changed

  • Fixed Docker-based Hermes integration for installs where the Hermes launcher or venv uses absolute paths.
  • Added configurable Hermes host/container mount paths: VW_HERMES_HOST_HOME, VW_HERMES_HOST_BIN, VW_HERMES_HOST_UV, and VW_HERMES_CONTAINER_UV.
  • Updated default Hermes binary path to the mounted launcher path so containers do not accidentally run the source-tree launcher with missing Python dependencies.
  • Bumped package version to 1.0.16.

Verification

  • Ran npm test / verify:smoke successfully.
  • Verified live Hermes chat after the fix returned OK on Eli's deployment.
  • Verified Hermes approval cards render in the Virtual World chat UI on the Aster-capable instance.

My Virtual World v1.0.15

Choose a tag to compare

@eliautobot eliautobot released this 16 Jun 06:26

Highlights

  • Adds the edit-panel + New Agent flow for creating agents directly inside My Virtual World.
  • Adds backend-driven platform discovery through /api/agent-platforms.
  • Adds /api/agent/create support for OpenClaw, Hermes, and Codex agents.
  • Adds native Codex provider support with configurable Codex home, workspace roots, sandbox, approval policy, and native agent registration.
  • Expands Hermes provider support to discover, create, and delete profile-backed product agents.
  • Adds location fields under Settings -> General and the setup wizard for future location-aware Day & Time Cycle, Time, and Weather behavior.
  • Updates Docker/env docs for OpenClaw host paths, Hermes launcher path, and Codex CLI mounts/settings.

Verification

  • Local npm test passed.
  • Local git diff --check passed.
  • Live 8590 /healthz passed before release.
  • GitHub Actions Smoke Test passed for the release commit bac9198.
  • New Agent create/delete was tested for OpenClaw, Hermes, and Codex before the release update.

v1.0.14

Choose a tag to compare

@eliautobot eliautobot released this 15 Jun 19:35
  • Fixes chat window agent pickers so saved or opened chats keep the correct agent instead of falling back to Adam.
  • Raises the mini chat bubble above agent name labels so it no longer covers the name.
  • Keeps carried desk drinks and snacks on the desk while agents are working, with hand carry only during active eating or drinking animations.
  • Adds smoke coverage for the chat picker, chat bubble cache key, and desk carry visual rules.
  • Bumps the package version to 1.0.14.