Daemon Hot Query Cache

Warm daemon queries under 6 ms

Benchmark loop against a Linux kernel checkout with explicit daemon/local equivalence checks and full Rust test coverage.

455.37 msbaseline daemon p95
4.90 msbest retained p95
20 / 20daemon/local hits
7 / 0equivalence cases/failures

Main Optimization Loop

Main daemon hot-query benchmark plot
StepChangedaemon_hot_p95_msDelta
Baselineprocess-cold CLI vs warm daemon baseline455.37-
1cache daemon SearchContext by workspace and embedding dimension240.04-215.32
2cache repeated daemon query results behind bounded request/index key219.03-21.01
3use quick query index health before daemon hot queries8.04-211.00

Final manual validation after cleanup measured daemon_hot_p95_ms = 5.88 with daemon_hits = 20 and local_hits = 20.

Raw main-loop TSV

Regression Exploration Loop

Explore daemon hot-query benchmark plot
StepChangedaemon_hot_p95_msDelta
Baselinepost-main-loop 9-sample baseline6.73-
1skip daemon Status preflight for static hot-query runs with existing socket4.90-1.83

The explore guard kept daemon/local JSON equivalence explicit: 7 representative cases, 0 failures, including --all after single-workspace cache warmup.

Raw exploration TSV

Validation