---

name: websocket-toolkit-googleai-dart description: Automates updating googleai_dart when Gemini Live API WebSocket schema changes. Fetches latest schema, compares against current, generates changelogs and prioritized implementation plans. Use for: (1) Checking for Live API updates, (2) Generating implementation plans for WebSocket changes, (3) Creating new message types from schema, (4) Syncing local schema with upstream. Triggers: "update live api", "sync websocket", "new messages", "live api changes", "check for live updates", "update live schema", "websocket version", "fetch live schema", "compare live schema", "what changed in live api", "live implementation plan".

WebSocket Toolkit (googleai_dart)

Uses shared scripts from websocket-toolkit with googleai_dart-specific configuration.

Uses verification scripts from openapi-toolkit.

Prerequisites

• GEMINI_API_KEY or GOOGLE_AI_API_KEY environment variable set
• Working directory: Repository root
• Python 3

Spec Registry

Spec	Description	Auth Required
live	Gemini Live API (real-time audio/video/text streaming)	Yes

Workflow

1. Fetch Latest Schema

python3 .claude/shared/websocket-toolkit/scripts/fetch_schema.py \
  --config-dir .claude/skills/websocket-toolkit-googleai-dart/config

Output: /tmp/websocket-toolkit-googleai-dart/latest-live.json

2. Analyze Changes

python3 .claude/shared/websocket-toolkit/scripts/analyze_changes.py \
  --config-dir .claude/skills/websocket-toolkit-googleai-dart/config \
  packages/googleai_dart/specs/live-api-schema.json /tmp/websocket-toolkit-googleai-dart/latest-live.json \
  --format all \
  --changelog-out /tmp/websocket-toolkit-googleai-dart/changelog-live.md \
  --plan-out /tmp/websocket-toolkit-googleai-dart/plan-live.md

Generates:

• changelog-live.md - Human-readable change summary
• plan-live.md - Prioritized implementation plan (P0-P4)

3. Implement Changes

Before implementing, read references/implementation-patterns.md for:

• Sealed class structure for messages
• WebSocket connection patterns
• JSON serialization for WebSocket messages

Use templates from ../../shared/websocket-toolkit/assets/:

• sealed_message_template.dart - Sealed class for WebSocket messages
• model_template.dart - Model class structure
• test_template.dart - Unit test structure

3.5 Update Documentation (MANDATORY)

Before running the review checklist:

1. README.md - Add/update Live API section
2. example/ - Create/update live_example.dart
3. CHANGELOG.md - Add entry for new features

4. Review & Validate (MANDATORY)

# Pass 2: Barrel file verification (from shared/openapi-toolkit)
python3 .claude/shared/openapi-toolkit/scripts/verify_exports.py \
  --config-dir .claude/skills/websocket-toolkit-googleai-dart/config

# Pass 3: Documentation completeness
python3 .claude/shared/openapi-toolkit/scripts/verify_readme.py \
  --config-dir .claude/skills/websocket-toolkit-googleai-dart/config
python3 .claude/shared/openapi-toolkit/scripts/verify_examples.py \
  --config-dir .claude/skills/websocket-toolkit-googleai-dart/config
python3 .claude/shared/openapi-toolkit/scripts/verify_readme_code.py \
  --config-dir .claude/skills/websocket-toolkit-googleai-dart/config

# Pass 4: Property-level verification
python3 .claude/shared/openapi-toolkit/scripts/verify_model_properties.py \
  --config-dir .claude/skills/websocket-toolkit-googleai-dart/config

# Dart quality checks (run from packages/googleai_dart)
cd packages/googleai_dart && dart analyze --fatal-infos && dart format --set-exit-if-changed . && dart test test/unit/

5. Testing (MANDATORY)

Test locations:

• Config classes: test/unit/models/live/config/
• Message types: test/unit/models/live/messages/
• Enums: test/unit/models/live/enums/

cd packages/googleai_dart && dart test test/unit/models/live/

6. Finalize

cp /tmp/websocket-toolkit-googleai-dart/latest-live.json packages/googleai_dart/specs/live-api-schema.json
cd packages/googleai_dart && dart test && dart analyze && dart format --set-exit-if-changed .

WebSocket Endpoints

Google AI:

wss://generativelanguage.googleapis.com/v1beta/models/{model}:BidiGenerateContent?key={API_KEY}&alt=ws

Vertex AI:

wss://{location}-aiplatform.googleapis.com/ws/google.cloud.aiplatform.v1beta1.PredictionService.BidiGenerateContent
Authorization: Bearer {ACCESS_TOKEN}

Message Types

Client Messages

• BidiGenerateContentSetup - Initial session configuration
• BidiGenerateContentClientContent - User content/context
• BidiGenerateContentRealtimeInput - Real-time audio/video/text input
• BidiGenerateContentToolResponse - Tool execution responses

Server Messages

• BidiGenerateContentSetupComplete - Session ready confirmation
• BidiGenerateContentServerContent - Model responses
• BidiGenerateContentToolCall - Tool execution requests
• BidiGenerateContentToolCallCancellation - Tool call cancellations
• GoAway - Session ending notification
• SessionResumptionUpdate - Resumption token updates

Package-Specific References

• Implementation Patterns - WebSocket patterns
• Live API Schema - Schema documentation
• Review Checklist - Validation process

Troubleshooting

• API key error: Export GEMINI_API_KEY or GOOGLE_AI_API_KEY
• Network errors: Check connectivity; retry after a few seconds
• No changes detected: Summary shows all zeros; no action needed