Skip to content

MODERATION_CONFIG_CACHE.md

Latest commit

 

History

History
75 lines (54 loc) · 2.63 KB

MODERATION_CONFIG_CACHE.md

File metadata and controls

75 lines (54 loc) · 2.63 KB
Raw

Moderation Config Cache

The moderation module loads its runtime settings into one immutable in-memory snapshot. A moderation check reads that snapshot once and shares it across keyword filtering, score thresholds, image checks, and the optional external moderation APIs.

Cached Settings

  • blockOnHighSeverity
  • highThreshold
  • mediumThreshold
  • external text and image API URLs and keys
  • blocked keywords
  • blocked image hosts

The first check loads and normalizes the moderation configuration namespace. Later checks reuse the same object without repeating source lookups.

Invalidation

Runtime settings writers must use ModerationConfigCacheService.updateConfig(update). The method updates Nest's configuration source, invalidates the existing snapshot, and eagerly returns the newly loaded snapshot. invalidate() is also available when another configuration writer owns persistence.

Each application process owns its cache. A future database-backed or multi-process settings writer must broadcast an invalidation event to every process after a successful update.

Metrics

The existing metrics endpoint exports:

  • moderation_config_source_loads_total
  • moderation_config_cache_invalidations_total
  • moderation_config_cache_load_duration_ms

These metrics make source reloads and invalidations visible without adding work to the cache-hit path or introducing high-cardinality values.

Benchmark

Run the deterministic microbenchmark from BackEnd:

npm run benchmark:moderation-config

Override the default one million iterations when needed:

MODERATION_CONFIG_BENCHMARK_ITERATIONS=2000000 \
  npm run benchmark:moderation-config

The benchmark compares the previous three ConfigService.get() threshold reads per moderation decision with one cached snapshot lookup. It reports elapsed time, throughput, and source-read counts as JSON.

Recorded Baseline

Measured on July 27, 2026 with Node.js v24.16.0, using the default 1,000,000 iterations and the median of five interleaved runs:

Measurement Before After Change
Elapsed time 4,021.48 ms 140.90 ms 96.50% lower
Throughput 248,665 decisions/s 7,097,408 decisions/s 28.54x
Config source reads 3,000,000 1 99.99997% lower

This microbenchmark uses the current Nest ConfigService source. A database-backed source would retain the same one-load cache behavior while avoiding the additional database round-trip cost.