parallel-agent-spawner（平行 Agent 生成器）

Parallel Agent Spawner

Spawns and coordinates parallel agents for faster feature completion.

Quick Start

Spawn Parallel Agents

from scripts.parallel_spawner import ParallelSpawner

spawner = ParallelSpawner(project_dir)
results = await spawner.spawn_parallel(
    tasks=["feature-1", "feature-2", "feature-3"],
    agent_type="coding"
)

Coordinate Results

await spawner.wait_all()
summary = spawner.get_results_summary()

Parallel Execution Model

┌─────────────────────────────────────────────────────────────┐
│                   PARALLEL EXECUTION                        │
├─────────────────────────────────────────────────────────────┤
│                                                             │
│  MAIN ORCHESTRATOR                                          │
│  ├─ Analyzes feature dependencies                          │
│  ├─ Identifies parallelizable work                         │
│  ├─ Spawns worker agents                                   │
│  └─ Coordinates results                                    │
│                                                             │
│  ┌──────────┐  ┌──────────┐  ┌──────────┐                 │
│  │ Agent 1  │  │ Agent 2  │  │ Agent 3  │                 │
│  │          │  │          │  │          │                 │
│  │ feature-1│  │ feature-2│  │ feature-3│                 │
│  └────┬─────┘  └────┬─────┘  └────┬─────┘                 │
│       │             │             │                        │
│       ▼             ▼             ▼                        │
│  ┌─────────────────────────────────────────┐              │
│  │           RESULT AGGREGATOR              │              │
│  │  ├─ Collects all results                │              │
│  │  ├─ Resolves conflicts                  │              │
│  │  ├─ Merges code changes                 │              │
│  │  └─ Updates feature list                │              │
│  └─────────────────────────────────────────┘              │
│                                                             │
└─────────────────────────────────────────────────────────────┘

Agent Types

Spawn Configuration

@dataclass
class SpawnConfig:
    max_parallel: int = 3
    timeout_per_agent: int = 1800  # 30 min
    retry_failed: bool = True
    merge_strategy: str = "sequential"  # or "git-merge"

Integration Points

• autonomous-loop: Provides work to parallelize
• coding-agent: Worker agents for features
• checkpoint-manager: Create checkpoints before spawn
• error-recoverer: Handle agent failures

References

• references/PARALLELIZATION-STRATEGY.md - Strategy guide
• references/CONFLICT-RESOLUTION.md - Merge conflicts

Scripts

• scripts/parallel_spawner.py - Core spawner
• scripts/agent_pool.py - Agent pool management
• scripts/result_aggregator.py - Aggregate results
• scripts/dependency_analyzer.py - Find parallelizable work