Add benchmarks and improvements for contextSelector#2644

Open

GeorgeNgMsft wants to merge 22 commits into

mainfrom

dev/georgeng/benchmark_contextselector

Commits on Jul 10, 2026

Add benchmark harness and report for context selector
GeorgeNgMsft
committed

Commits on Jul 14, 2026

Merge branch 'main' into dev/georgeng/benchmark_contextselector
GeorgeNgMsft
authored
Add contextSelector strategy comparison, run-all harness, and test suites

GeorgeNgMsft
and
Copilot
committed
Consolidate contextSelector benchmark output into one local report

GeorgeNgMsft
and
Copilot
committed
Clarify contextSelector report: split spurious/wrong-target, add %s, define retrieval

GeorgeNgMsft
and
Copilot
committed
Rewrite contextSelector report for newcomers: overview, key terms, metric goals

GeorgeNgMsft
and
Copilot
committed
Add executive summary to the top of the contextSelector report

GeorgeNgMsft
and
Copilot
committed
Update report generation
GeorgeNgMsft
committed
Merge branch 'dev/georgeng/benchmark_contextselector' of https://github.com/microsoft/TypeAgent into dev/georgeng/benchmark_contextselector
GeorgeNgMsft
committed
Document benchmark structure: 3 layers, fixture shape, per-generator examples

GeorgeNgMsft
and
Copilot
committed
Default the benchmark report to the tracked docs copy and auto-format it

GeorgeNgMsft
and
Copilot
committed
Add fidelity caveat to the contextSelector-vs-LLM report section

GeorgeNgMsft
and
Copilot
committed
Enable contextSelector collision resolution by default

GeorgeNgMsft
and
Copilot
committed