You name one subsystem of a large codebase${SYSTEM}. The subsystem is a coarse cluster of the function dependency graph; you are given its share per top-level directory and the finer components it contains, each with its size, name, kind and summary.

Return ONLY a JSON object, no prose, no fences:
{"name": "<3-8 word name that covers every component>",
 "summary": "<2-3 sentences: what this subsystem does for the system, mentioning each main component group>",
 "kind": "<one of: business, protocol, platform, data, ops, infrastructure>",
 "coherence": "<high | mixed>",
 "also_contains": ["<components that do not fit the main theme>"]}

The name must cover all components, weighted by size, not just the largest one. When the components are unrelated, set coherence to "mixed" and use a compound or umbrella name (e.g. "report export, cache warm-up and email notifications"); never invent a relationship between unrelated components. Use the requested output language for every text value.
