Hi, I've tried to implemented this against MongoDB Search Community (mongot): https://github.com/TinTinWinata/mongot/pull/1/.

Actually, this isn't the per-bucket feature this idea asks for. What I've built is the global case, metrics over the entire match set with no bucketing. But I tried to do in facet as well, and it's works, i just didn't have time to finalize it, because currently i don't need that use cases.

Example of how i use it:

db.readings.aggregate([{ $searchMeta: {
  index: "readingsIndex",
  metrics: {
    operator: { compound: { filter: [{ equals: { path: "reading.category", value: 0 } }] } },
    metrics: {
      minValue: { type: "min", path: "reading.value" },
      maxValue: { type: "max", path: "reading.value" },
      avgValue: { type: "avg", path: "reading.value" }
    }
  }
}}])
// => { count: { lowerBound: 128431 }, metrics: { minValue: 0.5, maxValue: 132.0, avgValue: 41.7 } }

If it's easier to poke at than to read, there's a prebuilt image of the fork:

docker pull tintin6892/mongot-community:searchmeta-metrics-9d37d75a