docs: release v0.7.4 with portable setup and accurate workload guides

Remove deployment-specific defaults and observations from public guidance.
Require explicit source-sync destinations and resolve pg-stand from the active
Python environment. Document the active HTAP and IMDb query sets and index
the source synchronization guide.

Validation: 457 tests passed on Python 3.10 and 3.12; Ruff, wheel/sdist,
27 packaged guides, and fresh-wheel CLI smoke checks passed.

Author: O2eg

doc/README.md

@@ -18,6 +18,7 @@ initializer documentation, and packaged profile and JOIN READMEs.
 | [SSH transport](ssh_mode_usage.md) | PostgreSQL runs on a remote host reached through SSH |
 | [Managed PostgreSQL](managed_mode_usage.md) | SQL-only access to a pre-created benchmark database |
 | [Workload configuration](workload_description.md) | You are defining the init command, measured command, and iteration axis |
+| [Source synchronization](sidecar-source-workflow.md) | You are developing from a checkout and updating a separate load-generator host |
 | [Reports and comparisons](logic_building_and_comparing_reports.md) | You are consuming JSON/HTML artifacts or joining benchmark runs |
 | [`pg_play` integration](pg_play-integration.md) | You are invoking the versioned machine contract from an orchestrator |
 

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "pg-perf-bench"
-version = "0.7.3"
+version = "0.7.4"
 description = "Reproducible PostgreSQL TPS benchmarks with environment evidence and report comparison"
 readme = "README.md"
 requires-python = ">=3.10"

src/pg_perf_bench/__init__.py

@@ -1,3 +1,3 @@
 """PostgreSQL performance benchmark and evidence collector."""
 
-__version__ = '0.7.3'
+__version__ = '0.7.4'

src/pg_perf_bench/workload_profiles/imdb/sql/planner/README.md

@@ -27,8 +27,9 @@ implementation or a realistic traffic mix.
 
 ## Choosing a workload
 
-Use the 14 scripts in the parent `sql` directory for the supported catalog workload.
-They return coherent records or explicitly defined aggregate groups and are
+Use the 11 active scripts in the parent `sql` directory for the default catalog workload;
+three other scripts are disabled and excluded from the default command.
+The active scripts return coherent records or explicitly defined aggregate groups and are
 validated against an independent source-record oracle. See the
 [profile README](../../README.md) for data contracts, weights and timeout policy.
 

src/pg_perf_bench/workload_profiles/pagila-htap/README.md

@@ -7,9 +7,12 @@ current logical workload day. All generated receipt dates are reachable; the upp
 window bound advances with `benchmark_now()` without rewriting a shared counter. All period boundaries
 and daily groups use UTC, independently of the server's `TimeZone`.
 
-The seven dashboard queries return cashier revenue, category revenue, catalog
-counts for the selected category, daily rentals of that category, its most-rented
-titles, daily cash receipts, and loyalty candidates. Revenue is attributed to
+Four dashboard queries are active by default: category catalog counts (query 3),
+most-rented titles (5), daily cash receipts (6), and loyalty candidates (7).
+Cashier revenue (1), category revenue (2), and daily rentals/returns (4) remain
+in block comments pending optimization and are not executed. To include one,
+uncomment its block in a copied custom profile and use the same SQL on all
+compared environments. Revenue is attributed to
 payment dates; `paid_rentals` counts distinct rentals within a payment day.
 Rentals are counted directly from `rental`, so extra payments do not inflate them.
 Store attribution follows the issuing cashier's store, not a copy's current
@@ -27,7 +30,7 @@ are byte-identical to `pagila`. Weights remain `01_select` 50, `02_insert` 25,
 `03_update` 20, `04_delete` 5, `05_reporting` 5. Reporting probability is 5/105
 (about 4.76%); its fraction of elapsed client time depends on query latency.
 Clients share this mix, so the profile does not reserve a separate analyst pool.
-The seven statements run in a read-only transaction at the connection's isolation
+The four active statements run in a read-only transaction at the connection's isolation
 level; default READ COMMITTED does not promise a single cross-statement snapshot.
 
 The shared [data and time contract](../pagila/README.md#data-and-time-contract)