perf: reduce detection overhead from ~1.30x to ~1.03x#52

Merged

taobojlen merged 15 commits intomainfrom

perf/autoresearch-optimizations

Mar 18, 2026

Commits on Mar 18, 2026

perf: replace inspect.stack() with sys._getframe() in notify fast path → overhead_ratio=1.06

taobojlen
and
claude
committed
perf: cache allowlisted (model, field) pairs to skip redundant _alert() calls → overhead_ratio=1.05

taobojlen
and
claude
committed
perf: reduce per-call allocations in notify() and queryset patching → overhead_ratio=1.04

taobojlen
and
claude
committed
docs: update autoresearch.md progress log with iteration 3 results

taobojlen
and
claude
committed
perf: replace inspect.stack() with sys._getframe() in SHOW_ALL_CALLERS path → overhead_ratio=1.05, overhead_ratio_allcallers=1.07

taobojlen
and
claude
committed
docs: update autoresearch.md progress log with iteration 4 results

taobojlen
and
claude
committed
perf: replace pattern list iteration with direct substring checks in frame filtering → overhead_ratio=1.05, overhead_ratio_allcallers=1.06

taobojlen
and
claude
committed
docs: update autoresearch.md progress log with iteration 5 results

taobojlen
and
claude
committed
perf: lazy-cache settings lookups in notify() to eliminate per-call hasattr() overhead → overhead_ratio=1.04, overhead_ratio_allcallers=1.03

taobojlen
and
claude
committed
docs: update autoresearch.md progress log with iteration 6 results

taobojlen
and
claude
committed
docs: record noise floor analysis in autoresearch progress log (iteration 7)

taobojlen
and
claude
committed
perf: update README overhead claim and add autoresearch infrastructure

taobojlen
and
claude
committed
refactor: clean up util.py — remove dead code, rename fast-path functions

taobojlen
and
claude
committed
fix: use actual package directory for frame filtering instead of "/zeal/" substring

taobojlen
and
claude
committed
test: add tests for _is_internal_frame to verify frame filtering

taobojlen
and
claude
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

perf: reduce detection overhead from ~1.30x to ~1.03x#52

perf: reduce detection overhead from ~1.30x to ~1.03x#52
taobojlen merged 15 commits intomainfrom
perf/autoresearch-optimizations

Commits on Mar 18, 2026

perf: replace inspect.stack() with sys._getframe() in notify fast path → overhead_ratio=1.06

perf: cache allowlisted (model, field) pairs to skip redundant _alert() calls → overhead_ratio=1.05

perf: reduce per-call allocations in notify() and queryset patching → overhead_ratio=1.04

docs: update autoresearch.md progress log with iteration 3 results

perf: replace inspect.stack() with sys._getframe() in SHOW_ALL_CALLERS path → overhead_ratio=1.05, overhead_ratio_allcallers=1.07

docs: update autoresearch.md progress log with iteration 4 results

perf: replace pattern list iteration with direct substring checks in frame filtering → overhead_ratio=1.05, overhead_ratio_allcallers=1.06

docs: update autoresearch.md progress log with iteration 5 results

perf: lazy-cache settings lookups in notify() to eliminate per-call hasattr() overhead → overhead_ratio=1.04, overhead_ratio_allcallers=1.03

docs: update autoresearch.md progress log with iteration 6 results

docs: record noise floor analysis in autoresearch progress log (iteration 7)

perf: update README overhead claim and add autoresearch infrastructure

refactor: clean up util.py — remove dead code, rename fast-path functions

fix: use actual package directory for frame filtering instead of "/zeal/" substring

test: add tests for _is_internal_frame to verify frame filtering