docs(6205): inception and family comments drop record, milestone and date citations

inception states the legacy stamp and the retired rulebook keys as facts
about stored records; family keeps the repeat-trigger calibration numbers
without the date and record they were measured on, and drops the instance
that taught the per-kind comparison.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
2026-10-21 20:29:03 -04:00
co-authored by Claude Opus 5.5
parent 5f96545ae1
commit dd4bc234e9
2 changed files with 26 additions and 30 deletions
+14 -17
View File
@@ -1,4 +1,4 @@
"""Family canon's promotion engine (milestone 463 step 3).
"""Family canon's promotion engine.
A family idea moves between three states — candidate, canon, retired — and
every move is a decision with a reason, written to `family_decisions`. No
@@ -62,7 +62,7 @@ logger = logging.getLogger(__name__)
#
# These are what an evaluation is judged against, on every install. They live
# here — and in the tool docstrings that quote them — rather than in any
# instance's rulebook (rules 115, 119).
# instance's rulebook, so every install judges by the same text.
CRITERIA = (
{
@@ -97,13 +97,11 @@ CRITERIA_KEYS = tuple(c["key"] for c in CRITERIA)
TRIGGERS = ("citation", "repeat", "milestone")
# The repeat trigger's similarity floor. MEASURED, not guessed (2026-10-06,
# #4989): against a description of one pattern known to have been built four
# times in four projects, the records that implement it scored 0.79-0.82, and
# the best match in a project that never built it scored 0.65. A trigger only
# opens an evaluation, so a false positive costs one judgment, not a wrong
# promotion. Step 5 measures the cross-language threshold against known pairs
# and supersedes this.
# The repeat trigger's similarity floor. MEASURED, not guessed: against a
# description of one pattern known to have been built four times in four
# projects, the records that implement it scored 0.79-0.82, and the best match
# in a project that never built it scored 0.65. A trigger only opens an
# evaluation, so a false positive costs one judgment, not a wrong promotion.
REPEAT_THRESHOLD = 0.80
# Words that say a citation is the SOURCE of a pattern rather than background.
@@ -119,12 +117,11 @@ _LINEAGE = re.compile(
_LINEAGE_WINDOW = 80
# Note kinds the repeat trigger compares: recorded knowledge and shapes, not
# the to-do list — and each only against its own kind (#5130). A repeat is the
# same KIND of thing made twice: two shapes built, or two ideas written down.
# the to-do list — and each only against its own kind. A repeat is the same
# KIND of thing made twice: two shapes built, or two ideas written down.
# Across kinds the score stops meaning that. One chunk of a long narrative
# note that mentions a shape's vocabulary scores like the shape itself: a
# dated release log matched a signed-APK CI snippet at 0.83, above this
# trigger's floor, and opened the log as a candidate. The cross-kind relation
# note that mentions a shape's vocabulary scores like the shape itself — above
# this trigger's floor — and would open the note as a candidate. The cross-kind relation
# that IS real — code built from an idea written elsewhere — has its own
# routes: the citation trigger when the lineage is named, and the shape ledger
# once the idea is canon.
@@ -172,7 +169,7 @@ async def _topic(session, user_id: int, topic_id: int | None, *, rules: bool) ->
async def _snapshot(session, idea: FamilyIdea | None) -> dict | None:
"""The idea's state, as a decision's before/after. Slugs, never ids: an id
inside JSON cannot be remapped by a restore (the #3182 trap)."""
inside JSON cannot be remapped by a restore."""
if idea is None:
return None
return {
@@ -219,7 +216,7 @@ async def get_idea(user_id: int, note_id: int) -> dict | None:
"platforms": await _platform_slugs(session, note_id),
"topic": await _topic(session, user_id, idea.topic_id, rules=True),
"adoptions": counts,
# Brief: the history is read in bulk; the full rows are list_decisions (#5131).
# Brief: the history is read in bulk; the full rows are list_decisions.
"decisions": [d.to_brief() for d in decisions],
"undoable_decision_id": _undoable_id(decisions),
})
@@ -478,7 +475,7 @@ async def propose(
Idempotent: an existing idea comes back unchanged and nothing is logged —
a trigger firing on every edit of a record must not grow the log. Write-
gated on the note: an idea is state on someone's record (rule 78).
gated on the note: an idea is state on someone's record.
"""
if not (reason or "").strip():
raise ValueError("a proposal needs a reason — what makes this a family idea?")
+12 -13
View File
@@ -1,4 +1,4 @@
"""Project inception — what a project was decided to inherit (milestone 297).
"""Project inception — what a project was decided to inherit.
A project's inheritance is a decision, not a default. The record lives on
``projects.inception``::
@@ -14,14 +14,13 @@ A project's inheritance is a decision, not a default. The record lives on
}
NULL = undecided → enter_project asks. ``legacy`` is the migration's stamp on
projects that existed before the step did (inherit-all / no design system /
no seed), so the ask fires only for projects created after this shipped.
projects that predate inception records (inherit-all / no design system /
no seed), so the ask fires only for projects created since.
Rules are not a choice any more. ``exclude_always_on_rulebooks`` went with the
always-on tier (milestone 394), and ``subscribe_rulebooks`` went with
subscriptions (milestone 414): a rule in a rulebook is global and applies to
every project, and a project's own rules are written on it directly. Migration
0101 strips both keys from stored records.
Rules are not a choice: a rule in a rulebook is global and applies to every
project, and a project's own rules are written on it directly. Migration 0101
strips the retired ``exclude_always_on_rulebooks`` and ``subscribe_rulebooks``
keys from stored records.
The shape and its validator are pure; ``decide`` composes the existing
services — set_project_design_system and the standard Systems seed — checks
@@ -31,7 +30,7 @@ half-applied decision is re-runnable rather than recorded as done.
``current_defaults`` is what the enter_project ask shows: what binds today
if nobody decides.
``platforms`` (milestone 463) is which platforms the project IS — the answer
``platforms`` is which platforms the project IS — the answer
that decides which family ideas reach it. A list of catalog SLUGS, not ids:
the record is JSON, and a slug survives a backup restore onto an install whose
ids differ, where an id inside JSON would come back naming another platform.
@@ -54,7 +53,7 @@ CHOICE_KEYS = ("design_system_id", "seed_systems", "platforms")
def validate_inception(choices) -> str | None:
"""The structural error an inception ``choices`` object would earn, or
None. Pure and checked BEFORE any effect is applied: a decision either
applies whole or errors whole (the StrictArgs lesson, #2709).
applies whole or errors whole.
Accepts three keys, each optional: ``design_system_id`` an int or None,
``seed_systems`` a bool, ``platforms`` a list of slugs or None. Unknown
@@ -152,7 +151,7 @@ async def decide(
choices: dict | None,
via: str,
) -> dict:
"""Record a project's inception decision and apply it (milestone 297).
"""Record a project's inception decision and apply it.
Owner-only. Validates the choices (pure) and every target (readable)
first; then, each idempotent: point the project at the design system
@@ -243,8 +242,8 @@ def _platform_line(defaults: dict) -> str:
async def inception_ask(user_id: int, project_id: int) -> dict:
"""The enter_project ask for an undecided project (milestone 297) — the
sibling of the systems-bootstrap ask (#2683): the project's OWN current
"""The enter_project ask for an undecided project — the sibling of the
systems-bootstrap ask: the project's OWN current
defaults, what to ask the operator, and the exact call that answers it.
Fail-open: a hint must never break the call it rides on."""
try: