Decompose Call into focused internal components

[PratimMallick]

Goal

closes AND-1318

Call.kt had grown to ~2,350 lines and mixed many unrelated responsibilities (REST API calls, media, session/RTC lifecycle, reconnect state machine, connectivity/ICE monitoring, events, stats, rendering, teardown). This made it hard to read, test, and evolve.

This PR decomposes Call into focused internal collaborators while keeping Call as a thin, binary-compatible public facade — no public API changes. Call.kt goes from 2,351 to 1,043 lines.

Implementation

Structure

flowchart LR
    Call["<b>Call</b><br/>public facade<br/>(binary-compatible)"]

    Orch["<b>Orchestrators</b><br/>drive a flow across siblings<br/><br/>CallJoinCoordinator<br/>CallReconnector<br/>CallLifecycleManager"]

    Svc["<b>Services</b><br/>each owns one concern<br/><br/>CallApiClient · CallMediaManager<br/>CallEventManager · CallStatsReporter<br/>CallRenderer · SessionMonitor<br/>CallIceConnectionMonitor<br/>CallConnectivityMonitor"]

    Sess["<b>CallSessionManager</b><br/>single writer for<br/>the RTC session"]

    Seams["<b>Seams</b><br/>the only route back to<br/>the Call instance<br/><br/>ClientCallRegistry<br/>RtcSessionFactory<br/>MediaManagerFactory"]

    Call -->|constructs, delegates| Orch
    Call -->|constructs, delegates| Svc
    Orch <-->|direct sibling calls| Svc
    Orch -->|read / write| Sess
    Svc -->|read| Sess
    Orch -.->|needs the Call instance| Seams
    Svc -.-> Seams
    Seams -.->|implemented by| Call

Loading

The rules the diagram encodes:

• Ownership points one way. Call constructs every component and delegates its public API down to them. No component holds a Call reference, and none of them can reach back up except through a seam.
• Siblings call each other directly, rather than bouncing through the facade. Where a component is constructed before a collaborator exists (CallReconnector and CallLifecycleManager are built before CallState, because the event pipeline depends on them), the dependency is injected as a lazy provider.
• One writer for the session. CallSessionManager exposes the RTC session as a read-only StateFlow with setActiveSession() as the sole write path, so join, reconnect and teardown can read it but only one place mutates it.
• Three seams cover the handful of operations that genuinely need the call object itself. ClientCallRegistry registers and deregisters the call in the client's ringing/active slots, telecom and per-call cleanup; RtcSessionFactory and MediaManagerFactory keep the Call argument of RtcSession and MediaManagerImpl at the facade boundary. Call implements them as anonymous objects capturing this, so the call identity never leaks into a component.
• No dependency cycles. CallReconnector reads connectivity straight from the connection module rather than through CallConnectivityMonitor, which would otherwise close a ConnectivityMonitor → Reconnector → ConnectivityMonitor cycle.

Components

Component	Owns
CallApiClient	Coordinator REST wrappers
CallEventManager	events flow, subscriptions, dispatch
CallStatsReporter	Stats collection/reporting, statsReport / statLatencyHistory
CallRenderer	Renderer binding, visibility, screenshots, incoming media overrides
CallMediaManager	Peer-connection factory, media manager, audio pipeline, screen sharing
CallSessionManager	The RTC session, session ids, location, reconnect bookkeeping
CallIceConnectionMonitor	Publisher/subscriber ICE-restart monitoring
CallConnectivityMonitor	Network listener + leave-on-disconnect timeout
CallJoinCoordinator	Join flow, retry loop, session creation/connection
CallReconnector	FAST/REJOIN/MIGRATE state machine, failed-SFU tracking
CallLifecycleManager	Leave/end/cleanup, leave guard, destroyed flag
SessionMonitor	SFU signal/event observers and session monitoring

Call keeps its existing member names as delegating accessors, so RtcSession, CallState and existing callers keep working. Facade one-liners that existed only to route a call from one component to another (stopConnectionMonitors, stopStatsReporting, cancelSfuObservers, shutDownJobsGracefully, monitorSession, getFailedSfuIdsSnapshot) are removed now that siblings talk directly.

Two production bugs found while repairing the test suite

• CallMediaManager evaluated eglBase().eglBaseContext to build an argument for MediaManagerFactory.create, forcing a real EGL context before the factory (and its test hook) ran. Call owns both the context and the factory, so the parameter is dropped and the factory resolves it itself.
• The reconnect loop reads connectivity from the connection module, but the tests' network injection had been repointed at CallConnectivityMonitor. The loop therefore polled the real provider and stalled on isConnected() without consuming an attempt. Injecting at the module restores three reconnect tests that had been failing silently since the first decomposition commit.

Merge with develop

Reconciled develop's Call changes into the new components: typed SFU connect-failure recovery (SfuConnectionResult + SfuConnectFailureCause) now lives in CallJoinCoordinator, and the unitTestRtcSessionFactory hook replaces TestInstanceProvider.rtcSessionCreator.

Testing

• ./gradlew :stream-video-android-core:testDebugUnitTest — pass (978 tests, 0 failures, 50 skipped)
• ./gradlew :stream-video-android-core:apiCheck — pass (public API unchanged / binary-compatible)
• ./gradlew :stream-video-android-core:spotlessCheck — pass
• ./gradlew assembleDebug — pass (all modules incl. demo-app)

New unit tests cover each extracted component: CallApiClientTest, CallEventManagerTest, CallStatsReporterTest, CallRendererTest, CallMediaManagerTest, CallSessionManagerTest, CallIceConnectionMonitorTest, CallConnectivityMonitorTest, CallJoinCoordinatorTest, CallReconnectorTest.

Two existing tests were rebuilt rather than patched, because they depended on internals that no longer exist:

• JoinRecoverableFailureTest spied Call and reflectively repointed a field inside CallJoinCoordinator. It now constructs the coordinator directly with mocked collaborators and drives joinInternal, covering the same three SFU connect-failure causes.
• CallReconnectorTest now builds the reconnector with its injected siblings and a real CallSessionManager, so the session and reconnect bookkeeping behave as they do in production instead of being stubbed.

Test-only reflection seams (injectSession, injectMockNetwork) are consolidated into a single CallTestSeams.kt rather than duplicated across test classes, so no test-only API was added to Call.

☑️Contributor Checklist

General

• I have signed the Stream CLA (required)
• Assigned a person / code owner group (required)
• Thread with the PR link started in a respective Slack channel (required internally)
• PR targets the develop branch
• PR is linked to the GitHub issue it resolves

Code & documentation

• Changelog is updated with client-facing changes (N/A — no client-facing/API changes)
• New code is covered by unit tests
• Comparison screenshots added for visual changes (N/A — no UI change)
• Affected documentation updated (KDocs on every new component and seam)
• Tutorial starter kit updated
• Examples/guides starter kits updated (stream-video-examples)

☑️Reviewer Checklist

• XML sample runs & works
• Compose sample runs & works
• Tutorial starter kit
• Example starter kits work
• UI Changes correct (before & after images)
• Bugs validated (bugfixes)
• New feature tested and works
• Release notes and docs clearly describe changes
• All code we touched has new or updated KDocs
• Check the SDK Size Comparison table in the CI logs

🎉 GIF

Skipped — internal refactor with no user-facing change.

Made with Cursor

[github-actions]

PR checklist ✅

All required conditions are satisfied:

• Title length is OK (or ignored by label).
• At least one pr: label exists.
• Sections ### Goal, ### Implementation, and ### Testing are filled, or the PR is bot-authored.
• An issue is linked (Linear ticket or GitHub issue), or the PR is bot-authored.

🎉 Great job! This PR is ready for review.

[github-actions]

SDK Size Comparison 📏

SDK	Before	After	Difference	Status
stream-video-android-core	12.27 MB	12.29 MB	0.02 MB	🟢
stream-video-android-ui-xml	5.68 MB	5.70 MB	0.02 MB	🟢
stream-video-android-ui-compose	6.20 MB	6.19 MB	-0.02 MB	🚀

[coderabbitai]

Walkthrough

Changes

Call responsibility decomposition

Layer / File(s)	Summary
Call wiring and shared state stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/Call.kt, stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallSessionManager.kt, stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallEventManager.kt	Call now exposes collaborator-backed session, event, media, statistics, and lifecycle state, with initialization-order coverage updated.
REST and call operation façade stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallApiClient.kt, stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/Call.kt	Coordinator REST calls, member actions, lifecycle actions, recordings, broadcasts, reactions, permissions, and transcription APIs are routed through CallApiClient.
Media, rendering, and statistics ownership stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallMediaManager.kt, CallRenderer.kt, CallStatsReporter.kt	Media setup, audio levels, peer connection factory recreation, screen sharing, rendering, screenshots, incoming track controls, and periodic statistics are delegated to dedicated components.
Join, connectivity, and reconnection flows stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallJoinCoordinator.kt, CallReconnector.kt, CallConnectivityMonitor.kt, CallIceConnectionMonitor.kt	Join retries, SFU session creation, network recovery, ICE recovery, reconnect strategy escalation, migration, and failed-SFU tracking are moved into coordinators and monitors.
Lifecycle teardown and cleanup stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallLifecycleManager.kt, stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/Call.kt	Guarded leave, end-call handling, analytics, media shutdown, statistics stopping, session cleanup, and scope disposal are centralized in CallLifecycleManager.

Estimated code review effort: 4 (Complex) | ~60 minutes

Sequence Diagram(s)

sequenceDiagram
  participant Call
  participant CallJoinCoordinator
  participant RtcSession
  participant CallReconnector
  participant CallLifecycleManager

  Call->>CallJoinCoordinator: join()
  CallJoinCoordinator->>RtcSession: connect to SFU
  RtcSession-->>CallJoinCoordinator: connection result
  Call->>CallReconnector: reconnect(strategy, reason)
  CallReconnector->>CallJoinCoordinator: rejoin or migrate
  CallJoinCoordinator->>RtcSession: create replacement session
  CallReconnector-->>Call: connected or exhausted
  CallReconnector->>CallLifecycleManager: leave after exhausted recovery

Loading

Possibly related PRs

• GetStream/stream-video-android#1716: Modifies the same join, rejoin, and migration paths with analytics instrumentation.
• GetStream/stream-video-android#1740: Changes SFU join and reconnect timeout handling used by these flows.
• GetStream/stream-video-android#1741: Updates recoverable SFU connection failures in the same join and reconnect control flow.

Suggested labels: pr:internal

Suggested reviewers: rahul-lohra, aleksandar-apostolov

Poem

I’m a rabbit with delegates bright,
Sorting calls through day and night.
Join, reconnect, render, and leave,
Each has a burrow up its sleeve.
Less tangled code, more hops in flight! 🐇

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 23.81% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Title check	✅ Passed	The title clearly describes the primary change: decomposing the Call class into focused internal components.
Description check	✅ Passed	The description covers the goal, implementation, testing, checklists, issue reference, and explains that no UI changes apply.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch refactor/call-class-decomposition

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Actionable comments posted: 15

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In
`@stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/Call.kt`:
- Around line 413-420: Remove the duplicated join preflight logic from Call’s
public join flow, including analytics, permission observation/checking, and
guest-token waiting around the shown joinCoordinator.join call. Keep that setup
exclusively in CallJoinCoordinator.join(), leaving Call’s join method as a
direct delegation with the existing arguments.
- Around line 759-765: Update Call.shutDownJobsGracefully so teardown does not
immediately cancel scope while supervisorJob children are draining. Complete the
supervisor to reject new children, await existing children within an owned
structured client scope, then perform final scope cancellation; handle
cancellation and child failures deterministically without relying on the
detached UserScope launch.

In
`@stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallConnectivityMonitor.kt`:
- Around line 73-92: Update the disconnect handling around
leaveTimeoutAfterDisconnect so the timeout coroutine is created only when
clientImpl.leaveAfterDisconnectSeconds is positive. Treat zero or negative
values as disabled, avoiding delay(0) and the subsequent call.leave flow;
preserve the existing reconnection check and timeout behavior for positive
values.
- Line 40: Remove or redact call.id from the logger tags used by
CallConnectivityMonitor, CallIceConnectionMonitor, CallJoinCoordinator, and
CallReconnector. Update the logger initialization in
stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallConnectivityMonitor.kt:40-40,
CallIceConnectionMonitor.kt:35-35, CallJoinCoordinator.kt:52-52, and
CallReconnector.kt:69-69 while preserving non-sensitive call context such as
call.type.

In
`@stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallEventManager.kt`:
- Around line 77-95: The fireEvent method currently invokes listeners while
holding the subscriptions monitor and allows listener exceptions to abort
dispatch. Snapshot active subscriptions under synchronized(subscriptions),
release the lock, invoke each matching listener with per-listener failure
isolation so remaining callbacks continue, then always call
events.tryEmit(event) afterward.

In
`@stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallIceConnectionMonitor.kt`:
- Around line 69-83: Update startSubscriberMonitor to collect the session and
subscriber flows rather than reading call.session.value?.subscriber?.value once
before collecting iceState. Keep the monitor active when the subscriber is
initially null and automatically switch to observing each replacement
subscriber’s iceState, matching the publisher monitor’s flow-observation pattern
while preserving the existing restart and logging behavior.

In
`@stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallJoinCoordinator.kt`:
- Around line 163-184: Update the join flow in CallJoinCoordinator so a failed
join also resets the join-and-ring progress state. Handle the error on the outer
join result, not only inside the flatMap block, and call
state.toggleJoinAndRingProgress(false) while preserving the existing
ring-failure handling.

In
`@stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallLifecycleManager.kt`:
- Around line 63-80: Sanitize all raw call identifiers in component logging:
update CallLifecycleManager.leave and internalLeave to remove or hash call.cid
and call.id, and sanitize or remove call.id from logger tags in
CallLifecycleManager (39-39), CallMediaManager (47-47), CallRenderer (43-43),
and CallStatsReporter (35-35). Preserve the existing log context without
exposing call IDs or CIDs.
- Around line 107-118: Update the coroutine launched in the call-leave flow to
wrap the leave analytics, tracing, stats, and leave-event operations in a
try/finally block, with cleanup() in finally. Ensure cleanup always runs after
the leave markers are attempted, including when onCallLeave() throws or the
coroutine is cancelled.

In
`@stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallMediaManager.kt`:
- Around line 270-272: Update CallMediaManager.cleanup() to dispose the owned
peer-connection factory during final teardown, in addition to calling
mediaManager.cleanup(). Reuse the existing factory instance and disposal
mechanism already used during recreation, ensuring native WebRTC resources are
released.

In
`@stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallReconnector.kt`:
- Around line 254-271: Update the FAST reconnect failure handling in
CallReconnector’s ReconnectOutcome.Failed branch to consult
MAX_FAST_RECONNECT_ATTEMPTS and escalate to the REJOIN strategy once the
configured fast-retry limit is reached. Preserve the existing migration and
deadline escalation conditions, and apply the same logic to the corresponding
failure branch around the alternate referenced location.
- Around line 178-187: Update the reconnect loop in CallReconnector so the
network-availability wait is skipped when the active strategy is DISCONNECT,
allowing the server-requested disconnect path to execute offline and produce
SFU_DISCONNECT. Preserve the existing connectivity polling behavior for other
reconnect strategies.

In
`@stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallRenderer.kt`:
- Around line 156-178: Update takeScreenshot around suspendCancellableCoroutine
to register an invokeOnCancellation handler that asynchronously removes
screenshotSink through the same call.scope cleanup path used after the first
frame. Ensure cancellation cleanup is safe if no frame has arrived, while
preserving the existing first-frame removal behavior.

In
`@stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallStatsReporter.kt`:
- Around line 45-57: Update CallStatsReporter.start so the coroutine keeps the
initial reportingIntervalMs delay, then sends the first stats report immediately
upon entering the while (isActive) loop; retain the interval delay only for
subsequent reports.

In
`@stream-video-android-core/src/test/kotlin/io/getstream/video/android/core/CallFieldDeclarationOrderTest.kt`:
- Around line 53-67: Update CallFieldDeclarationOrderTest to assert that
eventManagerLine precedes eventsLine, in addition to the existing ordering
checks. Keep the requirement that both eventManager and events are declared
before state.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro

Run ID: ce0c99a0-1916-4ef7-85a5-2b59619e7d82

📥 Commits

Reviewing files that changed from the base of the PR and between be52131 and 16b0487.

📒 Files selected for processing (17)

• stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/Call.kt
• stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallApiClient.kt
• stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallConnectivityMonitor.kt
• stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallEventManager.kt
• stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallIceConnectionMonitor.kt
• stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallJoinCoordinator.kt
• stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallLifecycleManager.kt
• stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallMediaManager.kt
• stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallReconnector.kt
• stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallRenderer.kt
• stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallSessionManager.kt
• stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/call/components/CallStatsReporter.kt
• stream-video-android-core/src/test/kotlin/io/getstream/video/android/core/CallFieldDeclarationOrderTest.kt
• stream-video-android-core/src/test/kotlin/io/getstream/video/android/core/reconnect/FailedSfuIdsTest.kt
• stream-video-android-core/src/test/kotlin/io/getstream/video/android/core/reconnect/ReconnectAttemptsCountTest.kt
• stream-video-android-core/src/test/kotlin/io/getstream/video/android/core/reconnect/ReconnectSessionIdTest.kt
• stream-video-android-core/src/test/kotlin/io/getstream/video/android/core/rtc/JoinRecoverableFailureTest.kt

[rahul-lohra]

Solid PR, please review the comments. They are mostly on code-quality. There is 1 related to usage of client scope

[rahul-lohra]

All queries are resolved now. Good work 👍

[sonarqubecloud]

Quality Gate failed

Failed conditions
75.0% Coverage on New Code (required ≥ 80%)

See analysis details on SonarQube Cloud

[stream-public-bot]

🚀 Available in v1.30.0