v0.2.4

• Send a process-level instance id on the /ws URL: the relay can now tell this process's own reconnects (old socket closes with 4001, ordinary reconnect) from a foreign session taking over the token (4000, fail-loud, unchanged). The id is generated once per process and survives reconnects and stop()/start() cycles.
• Reconnect backoff now resets to 1s after a successful connection instead of staying escalated to the 30s cap after one long outage.

JitPack: com.github.omnidebuglink:omnidebuglink_android:v0.2.4

v0.2.3 — Exit the host app on close code 4000

When the server closes the connection with code 4000 (the device token was just claimed by a newer connection), the SDK now terminates the host process (exitProcess(0)) after logging loudly, instead of staying stopped.

• Reconnecting after a takeover would ping-pong with the servers kicker forever — one token pair belongs to one device seat, so a kicked client must not come back.
• This is fail-loud by design: a token accidentally shipped in a release build can no longer keep a silent debug channel alive. Gate start() on BuildConfig.DEBUG or remove it from release builds.
• Normal brief disconnects are unaffected (watchdog-cancelled sockets can never re-enter onClosed; verified against OkHttp 4.12 internals).

Upgrade: bump to v0.2.3 — implementation 'com.github.omnidebuglink:omnidebuglink_android:v0.2.3' (JitPack).

v0.2.0

Breaking change

start() now takes only the device token — the relay URL (wss://api.omnidebuglink.dev/ws) is baked into the library. Self-hosted relays: change the exported constant and rebuild.

Constant: OmniDebugLink.RelayUrl.

OmniDebugLink.start(this, "<clientToken>")

JitPack: com.github.omnidebuglink:omnidebuglink_android:v0.2.0