Skip to content

Codebase intelligence for AI and humans: code health scores, auto-generated docs, git analytics, dead code detection, and architectural decisions via MCP.

fix(update): count onboarding pages in the generation progress total (#928)

Raghav Chamadiyaaad338da76e5
diff --git a/packages/cli/src/repowise/cli/commands/update_cmd/command.py b/packages/cli/src/repowise/cli/commands/update_cmd/command.py
index 7cf26cc84..04dd6270b 100644
--- a/packages/cli/src/repowise/cli/commands/update_cmd/command.py
+++ b/packages/cli/src/repowise/cli/commands/update_cmd/command.py
@@ -1123,14 +1123,21 @@ def run_update(
gen_progress.update(gen_task, total=total)
def _on_page_done(_page_id: str) -> None:
- nonlocal completed_pages
+ nonlocal completed_pages, total_pages
completed_pages += 1
+ # The total is an up-front estimate (see _announce_total); some page
+ # categories can't be predicted exactly, so keep it monotonic with the
+ # real count, so it never renders "43 of 41" (issue #922).
+ if total_pages is not None and completed_pages > total_pages:
+ total_pages = completed_pages
if emitter is not None:
emitter.page_done(
completed=completed_pages, total=total_pages, cost_usd=cost_tracker.session_cost
)
else:
- gen_progress.update(gen_task, advance=1, cost=cost_tracker.session_cost)
+ gen_progress.update(
+ gen_task, advance=1, total=total_pages, cost=cost_tracker.session_cost
+ )
# Checkpoint each page to the DB as it lands: a crash mid-generation used
# to lose every finished page (persist ran only at the very end), so the
@@ -1172,6 +1179,13 @@ def run_update(
emitter.error(str(exc))
raise
+ # Reconcile the bar to the pages actually produced. The up-front total
+ # is an estimate and onboarding slots may gate-skip, leaving it above
+ # the real count; snap total to completed so the bar reads N of N at
+ # 100% instead of stalling short (issue #922).
+ if gen_task is not None:
+ gen_progress.update(gen_task, total=completed_pages, completed=completed_pages)
+
# Surface the FAQ-weighted budget tilt when session demand shaped this run
# (silent when there is no history to weight; human console mode only).
if emitter is None and getattr(generator, "faq_demand_summary", None):
diff --git a/packages/core/src/repowise/core/generation/page_generator/orchestrate.py b/packages/core/src/repowise/core/generation/page_generator/orchestrate.py
index 6f6617011..e57ba2d5e 100644
--- a/packages/core/src/repowise/core/generation/page_generator/orchestrate.py
+++ b/packages/core/src/repowise/core/generation/page_generator/orchestrate.py
@@ -307,6 +307,19 @@ class _GenerationRun:
for l in self.kg_ctx.get_layers()
if len([n for n in l.get("nodeIds", []) if n.startswith("file:")]) >= 3
)
+ # Level-8 onboarding pages (the non-promoted slots) also emit, and
+ # were previously omitted here, which made the progress total read
+ # lower than the pages actually generated (issue #922: "43 of 41").
+ # Counted in full like every other category above; the completed-id
+ # subtraction below nets out any already-generated slot on resume.
+ # This is still an upper bound: a slot may gate-skip at generation
+ # time (build_context -> None), so the caller reconciles the bar to
+ # the real completed count when generation finishes.
+ onboarding_page_count = 0
+ if getattr(self.config, "enable_onboarding", True):
+ from .. import onboarding as _onboarding
+
+ onboarding_page_count = len(_onboarding.iter_specs())
estimated_total = (
counts["api_contract"]
+ counts["symbol_spotlight"]
@@ -317,6 +330,7 @@ class _GenerationRun:
+ int(self.selection.emit_repo_overview)
+ int(self.selection.emit_arch_diagram)
+ counts["infra_page"]
+ + onboarding_page_count
# Deterministic coverage tail also produces (zero-LLM) pages.
+ len(getattr(self.selection, "deterministic_tail_paths", []))
)
diff --git a/tests/unit/generation/test_announce_total.py b/tests/unit/generation/test_announce_total.py
new file mode 100644
index 000000000..589f44757
--- /dev/null
+++ b/tests/unit/generation/test_announce_total.py
@@ -0,0 +1,79 @@
+"""The progress total announced by ``_GenerationRun._announce_total``.
+
+Regression for issue #922: the up-front total omitted the level-8 onboarding
+pages (the non-promoted slots), so the bar could report more pages completed
+than its total (e.g. "43 of 41"). The estimate must now include one slot per
+registered onboarding spec that has not already been generated, gated by
+``config.enable_onboarding``.
+"""
+
+from __future__ import annotations
+
+from types import SimpleNamespace
+
+from repowise.core.generation import onboarding as _onboarding
+from repowise.core.generation.models import compute_page_id
+from repowise.core.generation.page_generator.orchestrate import _GenerationRun
+
+
+def _selection() -> SimpleNamespace:
+ """A selection with a single file page and nothing else, so the onboarding
+ contribution is the only moving part of the total."""
+ return SimpleNamespace(
+ counts=lambda: {
+ "api_contract": 0,
+ "symbol_spotlight": 0,
+ "file_page": 1,
+ "scc_page": 0,
+ "module_page": 0,
+ "infra_page": 0,
+ },
+ emit_repo_overview=False,
+ emit_arch_diagram=False,
+ deterministic_tail_paths=[],
+ )
+
+
+def _fake_run(*, enable_onboarding: bool, completed_ids: set[str]) -> SimpleNamespace:
+ announced: list[int] = []
+ return SimpleNamespace(
+ selection=_selection(),
+ kg_ctx=SimpleNamespace(available=False),
+ config=SimpleNamespace(enable_onboarding=enable_onboarding),
+ completed_ids=completed_ids,
+ on_total_known=announced.append,
+ job_system=None,
+ job_id=None,
+ _announced=announced,
+ )
+
+
+def test_total_includes_onboarding_pages() -> None:
+ specs = _onboarding.iter_specs()
+ assert specs, "expected onboarding subkinds to be registered on import"
+
+ run = _fake_run(enable_onboarding=True, completed_ids=set())
+ _GenerationRun._announce_total(run)
+
+ (total,) = run._announced
+ # 1 file page + one page per registered onboarding slot.
+ assert total == 1 + len(specs)
+
+
+def test_onboarding_disabled_excludes_those_pages() -> None:
+ run = _fake_run(enable_onboarding=False, completed_ids=set())
+ _GenerationRun._announce_total(run)
+
+ (total,) = run._announced
+ assert total == 1
+
+
+def test_already_generated_onboarding_pages_are_not_recounted() -> None:
+ specs = _onboarding.iter_specs()
+ done = {compute_page_id("onboarding", _onboarding.target_path(specs[0].slot))}
+ run = _fake_run(enable_onboarding=True, completed_ids=done)
+ _GenerationRun._announce_total(run)
+
+ (total,) = run._announced
+ # The one already-completed slot drops out of the remaining total.
+ assert total == 1 + len(specs) - 1