Files
FabledGrimoire/alembic/versions/0128_rule_enforcement.py
T
bvandeusenandClaude Opus 5.5 61077752b9
CI & Build / Plugin hooks (push) Successful in 14s
CI & Build / Python lint (push) Successful in 2s
CI & Build / TypeScript typecheck (push) Failing after 21s
CI & Build / Python tests (push) Successful in 1m55s
CI & Build / integration (push) Successful in 2m2s
CI & Build / Build & push image (push) Skipped
feat(7728): a rule declares its enforcement and salience
A rule now carries what it does when it arrives, declared by its author
rather than guessed from a score: enforce (the act is checked), hold (the
reply or act waits for one read) or inform (a line beside the work, the
default). And a salience, high, normal or low, for ranking among rules of
the same enforcement at a crowded moment. A preference is always inform.

Migration 0128 adds both with CHECKs. Every door carries them: the
service, the MCP create and update tools, REST, backup and restore, and
the rule editor, which also marks holding and enforced rules in the list.
Unknown values are refused rather than mapped. Nothing reads them yet
beyond the record; the ordering function and the lifecycle come next in
this step, and step 5 decides who holds.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-22 18:05:45 -05:00

49 lines
1.8 KiB
Python

"""rule enforcement and salience — what a rule does when it arrives, and how much it outranks
Revision ID: 0128
Revises: 0127
Create Date: 2026-11-22
Two declared columns on a rule, read by the one function that orders what
reaches a moment:
- `enforcement`: `enforce` (the act itself is checked), `hold` (the reply or act
waits for one read), `inform` (a line beside the work). Every existing rule
starts as `inform`; declaring which ones hold is the author's call.
- `salience`: `high`, `normal`, `low` — the author's ranking among rules of
the same enforcement.
"""
import sqlalchemy as sa
from alembic import op
revision = "0128"
down_revision = "0127"
branch_labels = None
depends_on = None
_ENFORCEMENTS = ("enforce", "hold", "inform")
_SALIENCES = ("high", "normal", "low")
def _in_list(column: str, values: tuple[str, ...]) -> str:
return f"{column} IN ({', '.join(repr(v) for v in values)})"
def upgrade() -> None:
op.add_column("rules", sa.Column("enforcement", sa.Text(), nullable=False, server_default="inform"))
op.add_column("rules", sa.Column("salience", sa.Text(), nullable=False, server_default="normal"))
op.create_check_constraint("ck_rules_enforcement", "rules", _in_list("enforcement", _ENFORCEMENTS))
op.create_check_constraint("ck_rules_salience", "rules", _in_list("salience", _SALIENCES))
# A preference claims no force: it is how work is liked done, not a bar.
op.create_check_constraint(
"ck_rules_preference_informs", "rules", "kind <> 'preference' OR enforcement = 'inform'",
)
def downgrade() -> None:
op.drop_constraint("ck_rules_preference_informs", "rules", type_="check")
op.drop_constraint("ck_rules_salience", "rules", type_="check")
op.drop_constraint("ck_rules_enforcement", "rules", type_="check")
op.drop_column("rules", "salience")
op.drop_column("rules", "enforcement")