neat/telemetry

Root bridge for NEAT telemetry, diagnostics, and export surfaces.

Telemetry is the evidence layer for a NEAT run. The controller is constantly making decisions about search pressure, speciation, lineage, novelty, complexity, and Pareto tradeoffs, but those decisions are hard to reason about if all you can see is a single best score. This chapter exists so a reader can learn how NeatapticTS turns that hidden runtime state into a readable stream of snapshots, summaries, and exportable artifacts.

The telemetry subtree has two complementary halves:

write path: recorder/, metrics/, and runtime/ build one generation snapshot, enrich it with derived signals, and store or stream it safely
read path: facade/, accessors/, and exports/ expose that captured state back to callers as inspection helpers, JSONL logs, and CSV reports

Read this chapter when you want to answer questions such as:

why is a run improving or stalling?
how much structural diversity is left in the population?
which species are growing, shrinking, or stagnating?
which objectives are active, and how did the Pareto frontier evolve?
what is the lightest export surface for notebooks, spreadsheets, or logs?

flowchart LR
  Loop["Evolution loop"] --> Recorder["recorder/<br/>shape one generation entry"]
  Recorder --> Metrics["metrics/<br/>attach diversity, lineage, objectives, perf"]
  Recorder --> Runtime["runtime/<br/>buffer, stream, trim safely"]
  Runtime --> Buffer["telemetry buffer<br/>and history archives"]
  Buffer --> Facade["facade/<br/>Neat-facing inspection helpers"]
  Buffer --> Accessors["accessors/<br/>small read-only snapshots"]
  Buffer --> Exports["exports/<br/>JSONL and CSV serialization"]