* fix: search NetSyncManager in all loaded scenes, not just active scene

When multiple scenes are loaded, GetDefaultServerDiscoveryPort() now
searches all loaded scenes for NetSyncManager if it's not found in
the active scene.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* refactor: simplify scene search loop in GetDefaultServerDiscoveryPort (#364)

* Initial plan

* refactor: simplify GetDefaultServerDiscoveryPort scene search loop

Co-authored-by: from2001 <387880+from2001@users.noreply.github.com>

---------

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: from2001 <387880+from2001@users.noreply.github.com>

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
Co-authored-by: from2001 <387880+from2001@users.noreply.github.com>

* feat: add POST /v1/rooms/{room_id}/global-variables endpoint

Add support for updating global variables per room via the FastAPI REST bridge, mirroring the existing client-variables endpoint.

- Add GlobalVarStore class for in-memory queued room-level global vars
- Add MAX_GLOBAL_VARS = 100 constant (matches server default)
- Add RoomBridge.apply_global_now_or_queue / _apply_global / flush_global_vars
- Update RoomBridge._loop to flush global vars when connected
- Add POST /v1/rooms/{room_id}/global-variables endpoint
- Add httpx dev dependency for FastAPI TestClient
- Add test_rest_bridge.py with 19 unit/endpoint tests

Closes #357

Co-authored-by: Masahiro Yamaguchi <from2001@users.noreply.github.com>

* fix: format battery drain rate for better readability

* feat: add pop method to GlobalVarStore and update flush_global_vars logic

* test: add unit tests for pop method in GlobalVarStore and flush_global_vars logic

* docs: update README to enhance REST bridge section and add global variables endpoint details

---------

Co-authored-by: claude[bot] <41898282+claude[bot]@users.noreply.github.com>
Co-authored-by: Masahiro Yamaguchi <from2001@users.noreply.github.com>

Update org name (psychic-vr-lab → styly-dev) and branch (main → develop)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

fix: correct broken GitHub link in server README

#361)

* fix: add automatic 1 Hz keepalive heartbeat for stealth Python clients

The Python stealth client was disconnecting because no periodic signal
was sent to keep the server's 5-second CLIENT_TIMEOUT from expiring.

Mirrors Unity's TransformSyncManager architecture:
- Add STEALTH_HEARTBEAT_INTERVAL = 1.0 (matches HEARTBEAT_INTERVAL_SECONDS)
- send_stealth_handshake() now sets _is_stealth_mode = True and starts
  the internal heartbeat via _send_stealth_heartbeat()
- _maybe_send_stealth_heartbeat() is called each _receive_loop iteration;
  it re-enqueues a stealth packet whenever the interval elapses
- No caller change needed: one send_stealth_handshake() call is enough

Also adds test_stealth_heartbeat.py with 11 tests covering the unit-level
timer logic and an integration test that verifies the client stays
registered past the server timeout with a short CLIENT_TIMEOUT config.

Closes #356

Co-authored-by: Masahiro Yamaguchi <from2001@users.noreply.github.com>

* fix: ensure stealth mode is activated only after successful heartbeat transmission

* fix: prevent client stop call on None client in integration tests

---------

Co-authored-by: claude[bot] <41898282+claude[bot]@users.noreply.github.com>
Co-authored-by: Masahiro Yamaguchi <from2001@users.noreply.github.com>

* feat: support multiple NICs for UDP discovery and connection management

* feat: implement support for multiple NICs in UDP discovery and connection management

* feat: optimize broadcast address handling in net_sync_manager and improve battery drain rate formatting in client simulator

* test: add unit tests for multi-NIC discovery and connection address handling

* fix: ensure single callback execution per discovery round in net_sync_manager

* fix: update editor version to 6000.0.70f1 in ProjectVersion.txt

* Update setting files

* feat: add --dealer-port and --pub-port CLI arguments

Add --dealer-port and --pub-port CLI arguments to styly-netsync-server,
consistent with the existing --server-discovery-port argument. CLI arguments
take precedence over config file values (CLI > config > defaults).

Closes #369

Co-authored-by: Masahiro Yamaguchi <from2001@users.noreply.github.com>

* Potential fix for pull request finding

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

* fix: load default configuration early for accurate help text

---------

Co-authored-by: claude[bot] <41898282+claude[bot]@users.noreply.github.com>
Co-authored-by: Masahiro Yamaguchi <from2001@users.noreply.github.com>
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

* feat: add rest_api_port configuration and CLI argument for REST API bridge

* feat: organize argument parser into groups and add version flag

* feat: update REST API port configuration and add tests for CLI argument overrides

* feat(client): add connection loss detection and reconnection

- Add on_connection_error EventHandler fired when connection is lost
- Add receive_timeout parameter (default 30s) for server-silence detection:
  silence exceeding the threshold triggers reconnect scheduling
- Add reconnect_delay parameter (default 5s) for time before reconnect attempt
- Add _trigger_reconnect(): resets client-ready state, stops discovery,
  fires on_connection_error, and schedules socket teardown/recreation
- Add _perform_socket_reconnect(): tears down and recreates ZMQ sockets,
  rebuilds the poller in the receive loop, and auto-restarts discovery
  if start_discovery() was previously called
- Propagate terminal ZMQ errors (ETERM/ENOTSOCK) from SUB/DEALER drain
  loops to the reconnect flow instead of silently continuing
- Store discovery port in start_discovery(); stop_discovery() clears it
  to prevent auto-restart (matching user intent)
- Add reconnect_count to stats dict
- Add 4 tests: event handler presence, constructor params, silence
  timeout triggering on_connection_error, and stats key

Closes #366

Co-authored-by: Masahiro Yamaguchi <from2001@users.noreply.github.com>

* fix(client): clear stale reconnect state and ensure proper discovery thread handling

* test: add test for resource cleanup on stop from connection error listener

* fix(client): adjust silence detection logic to prevent false-positive reconnects

---------

Co-authored-by: claude[bot] <41898282+claude[bot]@users.noreply.github.com>
Co-authored-by: Masahiro Yamaguchi <from2001@users.noreply.github.com>

…ct logic (#375)

* fix(net_sync_manager): improve discovery stop handling and auto-connect logic

* fix(client): address review issues in discovery auto-connect

- Add self-join guard in _stop_discovery_internal() to prevent
  RuntimeError when called from the discovery thread
- Only update server/dealer_port/sub_port when not already connected
- Retry discovery on start() failure instead of leaving client dead
- Add tests for auto-connect, self-join safety, and failure recovery

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* Rename code-review-netsync to review-netsync

* chore: remove obsolete code-review-netsync skill file

* docs: add Unity-Python feature parity guidelines to review-netsync skill

…ile descriptors (#378)

…removing source-NIC binding logic (#380)

* fix(server): update ROUTER identity on client reconnect to prevent RPC/NV message loss

When a client reconnects before the server timeout (e.g. after a
sleep/wake cycle), its DEALER socket is recreated with a new ZMQ
identity. The server was not updating the stored identity in the
existing-client branch, causing all ROUTER-based control messages
(RPC, NetworkVariable sync, device ID mapping) to be sent to the
stale identity and silently lost.

Fixes #381

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix(server): resync NV and ID mapping on client reconnect

When a client reconnects (e.g. after sleep/wake) before the server
timeout removes it, the identity update alone is not sufficient —
the client may have missed NetworkVariable changes and device ID
mapping updates during sleep. Detect identity changes as reconnection
events and trigger NV full sync and ID mapping re-broadcast to the
reconnected client.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix(unity): reset handshake and NV state on sleep/wake resume

On OnApplicationPause(false), reset _clientNo, _hasInvokedReady,
_shouldCheckReady, _shouldSendHandshake, and the NV initial sync
flag before calling StartNetworking(). Without this, stale state
from the previous session keeps IsReady true while the new
connection is still being established, which can cause RPCs to be
sent prematurely and NV initial sync to be skipped.

This mirrors the same reset logic already used in the connection
error handler and room switch paths.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix(tests): fix two Windows-specific test failures

- test_logging_cli: read log file with explicit utf-8 encoding to
  avoid cp932 decode errors on Windows
- test_timing_monotonic: use >= instead of > for monotonic time
  comparison, as Windows timer resolution can return equal values
  for consecutive calls with short sleep intervals

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* refactor(server): unicast NV resync on reconnect and add regression test

Address PR #382 review comments:

1. Refactor NV sync to extract payload-building helpers
   (_build_global_var_sync_payload, _build_client_var_sync_payload)
   and add _sync_network_variables_to_client() for targeted unicast.
   On reconnect, only the reconnecting client receives the NV snapshot
   instead of broadcasting to the entire room.

2. Add test_reconnect_identity.py covering the reconnect scenario:
   same device_id with new DEALER identity, verifying identity update
   and NV resync delivery.

3. Rename test_monotonic_time_always_increases to
   test_monotonic_time_never_decreases with updated docstring/print
   to match the relaxed >= assertion.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix(server): enhance NV cache management on connection loss and sleep/wake events

* Revert "fix(server): enhance NV cache management on connection loss and sleep/wake events"

This reverts commit 96bd820.

---------

Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-authored-by: Masahiro Yamaguchi <from2001@hotmail.com>

…sole messages (#384)

…playing (#385)

- Updated Unity package version in package.json
- Updated Unity version.txt file for runtime version retrieval
- Updated Python server version in pyproject.toml
- Updated version references in .md files (if any)