fix(benchmarks): address PR #76 review feedback from @timon0305

Drop unnecessary actions: write on benchmarks CI job; assert explicit
search hit count on list response; document 10x memory ceiling and v1
template limitations in README; note NoopSink export rounds are stateless.

Author: clean6378-max-it

.github/workflows/ci.yml

@@ -210,7 +210,6 @@ jobs:
     runs-on: ubuntu-latest
     permissions:
       contents: read
-      actions: write
     steps:
       - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
         with:

benchmarks/README.md

@@ -29,6 +29,10 @@ The memory test also runs as part of the normal `pytest` suite (timing benchmark
 
 Large JSONL files (5000+ lines) are generated at test session scope under pytest's temp directory — not committed to git.
 
+Corpora repeat one row from `tests/fixtures/session_with_tools.jsonl`, so parse/export numbers measure steady-state throughput on a narrow schema slice — not full parser branch coverage. Treat as v1 baselines, not exhaustive perf proof.
+
+The memory test (`test_parse_memory.py`) is intentionally **not** skipped by `--benchmark-skip`; it runs in the main `pytest` job and builds the session-scoped 5000-line fixture once per session.
+
 ## CI
 
 The `benchmarks` workflow job uploads `benchmark-results.json` as a downloadable artifact. There is no regression gate yet.

tests/benchmarks/test_export_bench.py

@@ -23,6 +23,7 @@ def test_bulk_export_session_count(
     projects = [{"name": "bench-project", "path": str(export_corpus), "display_name": "Bench"}]
 
     def _run() -> object:
+        # NoopSink + since="all" + empty last_export_sessions: no disk/state writes per round.
         return run_bulk_export(
             projects=projects,
             since="all",

tests/benchmarks/test_parse_memory.py

@@ -11,6 +11,9 @@
 def test_large_parse_peak_memory_under_ceiling(parse_large_file: Path) -> None:
     path = parse_large_file
     file_bytes = path.stat().st_size
+    # Issue #7 ceiling: Python heap peak (tracemalloc) vs on-disk JSONL size. Parsed
+    # dict/str objects often exceed raw bytes; 10x is a generous v1 guard — relax with
+    # a comment here if the parser legitimately grows.
     ceiling = file_bytes * 10
 
     tracemalloc.start()

tests/benchmarks/test_search_bench.py

@@ -16,4 +16,5 @@ def _run() -> object:
 
     resp = benchmark(_run)
     assert resp.status_code == 200
-    assert resp.get_json(), "expected search hits from synthetic searchable tokens"
+    hits = resp.get_json()
+    assert isinstance(hits, list) and len(hits) > 0, "expected search hits from synthetic corpus"