BenchmarkService class

Performance benchmark service

Constructors

BenchmarkService()

Properties

hashCode → int

The hash code for this object.

no setter, inherited

runtimeType → Type

A representation of the runtime type of the object.

no setter, inherited

Methods

noSuchMethod(Invocation invocation) → dynamic

Invoked when a nonexistent method or property is accessed.

inherited

toString() → String

A string representation of this object.

inherited

Operators

operator ==(Object other) → bool

The equality operator.

inherited

Static Methods

aggregateRoundStats(List<DetailedBenchmarkStats> rounds) → DetailedBenchmarkStats

Aggregate multiple round stats by flattening all measured samples.

benchmark(String name, Future<void> fn(), {int iterations = 10, required BenchmarkCategory category}) → Future<BenchmarkResult>

Run multiple iterations and measure avg/min/max

benchmarkDetailed(Future<void> fn(), {int warmupIterations = 5, int measuredIterations = 30}) → Future<DetailedBenchmarkStats>

Run detailed benchmark with warmup and reproducible summary.

benchmarkEmbed(String text, {int iterations = 10}) → Future<BenchmarkResult>

Embedding generation benchmark (ONNX-powered)

benchmarkHybridSearch(String dbPath, String queryText, List<double> queryEmbedding, int docCount, {int iterations = 20}) → Future<BenchmarkResult>

Hybrid Search benchmark

benchmarkIngestFfiEntrypoints({int targetBytes = 1 * 1024 * 1024, int embeddingDim = 384, int maxChunkChars = 1500, int overlapChars = 100, int batchSize = 16, String? restoreDbPath, String? dbPathOverride}) → Future<IngestFfiEntrypointBenchResult>

Measure FFI text-byte traffic across the three IngestSession entrypoints on a deterministic document: the canonical prepareSourceIngestion (Dart String), prepareSourceIngestionFromUtf8 (bytes), and prepareSourceIngestionFromFile (path-only). Stub embeddings keep the measurement focused on FFI byte traffic.

benchmarkIngestFfiTraffic({int targetBytes = 1 * 1024 * 1024, int embeddingDim = 384, int maxChunkChars = 1500, int overlapChars = 100, int batchSize = 16, String? restoreDbPath, String? dbPathOverride}) → Future<IngestFfiBenchResult>

Measure FFI text-byte traffic for an addDocument ingest, comparing the legacy chain (addSourceInCollection + chunker + addChunks) against the IngestSession chain (prepareSourceIngestion + take_embedding_batch + commit_embeddings) on a deterministic document.

benchmarkIngestHeap({int targetBytes = 4 * 1024 * 1024, int embeddingDim = 384, int maxChunkChars = 1500, int overlapChars = 0, int batchSize = 16, String? restoreDbPath, String? dbPathOverride}) → Future<IngestHeapBenchResult>

Measure peak process RSS during each ingest entrypoint variant on the realistic caller pattern:

benchmarkIngestLatency({int targetBytes = 1 * 1024 * 1024, int embeddingDim = 384, int maxChunkChars = 1500, int overlapChars = 0, int batchSize = 16, int warmupRuns = 2, int measuredRuns = 5, String? restoreDbPath, String? dbPathOverride}) → Future<IngestLatencyBenchResult>

Measure wall-clock latency of the full prepare → drain → finalize → delete cycle across the three ingest entrypoint variants. Stub embeddings (zeroed Float32List) keep the measurement focused on the FFI + Rust pipeline cost rather than ONNX inference time.

benchmarkSearch(String dbPath, List<double> queryEmbedding, int docCount, {int iterations = 20}) → Future<BenchmarkResult>

Search benchmark (Rust HNSW-powered)

benchmarkTokenize(String text, {int iterations = 50}) → Future<BenchmarkResult>

Tokenization benchmark (Rust-powered)

collectSamples(Future<void> fn(), {int warmupIterations = 5, int measuredIterations = 30}) → Future<List<double>>

Collect measured samples after warmup.

measureMs(Future<void> fn()) → Future<double>

Measure execution time of async code block

measureMsSync(void fn()) → double

Measure execution time of sync code block

runFullBenchmark({required String dbPath, dynamic onProgress(String)?}) → Future<List<BenchmarkResult>>

Run full benchmark suite

summarizeSamples(List<double> samplesMs, {required int warmupIterations}) → DetailedBenchmarkStats

Summarize measured samples with p50/p95/stddev.