feat(rulings): a command or edit touching an area's files shows its rulings, once per session (milestone 444 step 4, #4757)
CI & Build / Python lint (push) Successful in 3s
CI & Build / Plugin hooks (push) Successful in 12s
CI & Build / TypeScript typecheck (push) Successful in 54s
CI & Build / integration (push) Successful in 1m6s
CI & Build / Python tests (push) Failing after 1m22s
CI & Build / Build & push image (push) Skipped

A System's rulings (the Rulings section of its description) now reach the
work by path, not by similarity. Both PreToolUse arms resolve the files a
command or edit names to the Systems whose path_patterns cover them, and the
first touch in a session shows each area's rulings in one line; a repeat is
a one-line reference. A lookup, so no floor, no budget, no retrieval_logs row.

- services/system_rulings: parse_rulings, command_paths (reads and writes,
  relative to the repo root from any cwd; flags, URLs, globs skipped),
  rulings_for_paths
- /tool-rules takes root, cwd and seen_ruling_systems; /prior-art takes
  seen_ruling_systems; both return ruling_system_ids
- hooks share <sid>.rulings.ids (cleared on compaction by the ledger naming
  convention); the Bash hook sends the repo root and cwd
- system_usage_events (migration 0114): surfacings by source, pulls from
  get_system; carried by backup (v20) through the system map
- writing-records: rulings also arrive when the area's files are touched

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
2026-10-02 23:08:32 -04:00
co-authored by Claude Opus 5.5
parent a113c72b4f
commit 556872c039
17 changed files with 808 additions and 13 deletions
+58 -1
View File
@@ -13,6 +13,7 @@ from scribe.models.rule_version import RuleVersion
from scribe.models.design_system import DesignSystem, DesignToken
from scribe.models.note_usage import NoteUsageEvent
from scribe.models.rule_usage import RuleUsageEvent
from scribe.models.system_usage import SystemUsageEvent
from scribe.models.retrieval_tuning import RetrievalTuningEvent
from scribe.models.canonical_system import CanonicalSystem
from scribe.models.rulebook import RuleRelation, rule_systems as rule_systems_t
@@ -88,8 +89,12 @@ logger = logging.getLogger(__name__)
# v19 (2026-10) added lesson_no_rule (#4631): the "no rule fits" answer and its
# reason. Without it a restored lesson that was judged to stand alone reads as
# never judged, and lands back on the unjudged list.
# v20 (2026-10) added systems.path_patterns and system_usage_events (milestone
# 444): the files that are each area, and whether an area's rulings were read
# once shown. The usage rows restore through the SYSTEM map, for the reason
# the rule twin restores through the rule map.
# Bump when the serialized schema changes.
BACKUP_VERSION = 19
BACKUP_VERSION = 20
# Every table this backup carries, by its REAL name. Paired with _NOT_INCLUDED
# below, these two lists must together account for the entire schema — which is
@@ -131,6 +136,9 @@ _BACKED_UP = [
"lesson_rule_links",
# v19 (2026-10): "no rule fits" answers (#4631).
"lesson_no_rule",
# v20 (2026-10): System usage telemetry (milestone 444), for the reason
# its note and rule twins travel.
"system_usage_events",
]
# Tables intentionally NOT in the backup, surfaced in the payload so the gap is
@@ -231,6 +239,7 @@ _COLUMN_EXCLUSIONS: dict[str, set[str]] = {
"note_usage_events": {"id"},
# Same as the note twin: the surrogate key is re-issued on insert.
"rule_usage_events": {"id"},
"system_usage_events": {"id"},
# Same again — and everything else travels, because each remaining column
# is part of the argument: what moved, from what, to what, by whom, why.
"retrieval_tuning_events": {"id"},
@@ -319,6 +328,7 @@ _IMPORT_COLUMN_EXCLUSIONS: dict[str, set[str]] = {
"lesson_no_rule": set(),
"note_usage_events": {"id"},
"rule_usage_events": {"id"},
"system_usage_events": {"id"},
"retrieval_tuning_events": {"id"},
"design_systems": {
"id", "deleted_at", "deleted_batch_id", "created_at", "updated_at",
@@ -444,6 +454,17 @@ def _usage_event_rows(rows) -> list[dict]:
]
def _system_usage_event_rows(rows) -> list[dict]:
return [
{
"user_id": r.user_id, "system_id": r.system_id, "event": r.event,
"source": r.source, "project_id": r.project_id,
"created_at": r.created_at.isoformat() if r.created_at else None,
}
for r in rows
]
def _rule_usage_event_rows(rows) -> list[dict]:
return [
{
@@ -802,6 +823,9 @@ async def export_full_backup() -> dict:
rule_usage_events = (
await session.execute(select(RuleUsageEvent))
).scalars().all()
system_usage_events = (
await session.execute(select(SystemUsageEvent))
).scalars().all()
# Oldest first, so a restored history reads in the order the dials
# actually moved — the sequence IS the argument when a surface has been
# walked up and down.
@@ -854,6 +878,7 @@ async def export_full_backup() -> dict:
"design_tokens": _design_token_rows(design_tokens),
"note_usage_events": _usage_event_rows(usage_events),
"rule_usage_events": _rule_usage_event_rows(rule_usage_events),
"system_usage_events": _system_usage_event_rows(system_usage_events),
"retrieval_tuning_events": _retrieval_tuning_event_rows(
retrieval_tuning_events
),
@@ -991,6 +1016,12 @@ async def export_user_backup(user_id: int) -> dict:
rule_usage_events = (await session.execute(
select(RuleUsageEvent).where(RuleUsageEvent.rule_id.in_(_rule_ids))
)).scalars().all() if _rule_ids else []
# Scoped through the SYSTEM, for the reason the rule usage events
# above are scoped through the rule: `user_id` is who it fired for.
_system_ids = [sy.id for sy in systems]
system_usage_events = (await session.execute(
select(SystemUsageEvent).where(SystemUsageEvent.system_id.in_(_system_ids))
)).scalars().all() if _system_ids else []
# Scoped on user_id, and here that IS the right column — unlike the
# rule usage events directly above. These record changes to this user's
# OWN retrieval settings, which is what `user_id` means on this table;
@@ -1054,6 +1085,7 @@ async def export_user_backup(user_id: int) -> dict:
"design_tokens": _design_token_rows(design_tokens),
"note_usage_events": _usage_event_rows(usage_events),
"rule_usage_events": _rule_usage_event_rows(rule_usage_events),
"system_usage_events": _system_usage_event_rows(system_usage_events),
"retrieval_tuning_events": _retrieval_tuning_event_rows(
retrieval_tuning_events
),
@@ -1566,6 +1598,22 @@ def _build_rule_usage_event(row: dict, maps: _Maps) -> RuleUsageEvent | None:
)
def _build_system_usage_event(row: dict, maps: _Maps) -> SystemUsageEvent | None:
"""Resolved through the SYSTEM map — see `_build_rule_usage_event`."""
sid = maps.systems.get(row.get("system_id", 0))
if sid is None:
return None
return SystemUsageEvent(
user_id=maps.users.get(row.get("user_id") or 0),
system_id=sid,
event=row.get("event", ""),
source=row.get("source", ""),
project_id=(maps.projects.get(row["project_id"])
if row.get("project_id") else None),
created_at=_dt(row.get("created_at")),
)
def _build_repo_binding(row: dict, maps: _Maps) -> RepoBinding | None:
"""Small, but losing these means every bound repo quietly stops loading
its project at session start."""
@@ -1816,6 +1864,7 @@ async def _restore_v2(data: dict) -> dict:
"settings": 0, "rulebooks": 0, "rulebook_topics": 0, "rules": 0,
"systems": 0, "record_systems": 0, "design_systems": 0,
"design_tokens": 0, "note_usage_events": 0, "rule_usage_events": 0,
"system_usage_events": 0,
"repo_bindings": 0,
"note_supersessions": 0, "code_shapes": 0, "code_shape_events": 0,
"code_shape_uses": 0, "canonical_systems": 0,
@@ -2105,6 +2154,14 @@ async def _restore_v2(data: dict) -> dict:
session.add(event)
stats["rule_usage_events"] += 1
# And the System twin, after the Systems (step 16 fills their map).
for ev in data.get("system_usage_events", []):
event = _build_system_usage_event(ev, maps)
if event is None:
continue
session.add(event)
stats["system_usage_events"] += 1
# 20. Repo bindings
for rb_data in data.get("repo_bindings", []):
binding = _build_repo_binding(rb_data, maps)
+51 -6
View File
@@ -48,6 +48,7 @@ from scribe.services.retrieval_telemetry import record_retrieval
from scribe.services.settings import get_setting
from scribe.services.systems import system_names_for
from scribe.services.text import elide
from scribe.services import system_rulings as system_rulings_svc
logger = logging.getLogger(__name__)
@@ -2187,6 +2188,7 @@ async def build_write_path_hint(
exclude_derive: list[str] | None = None,
exclude_rule_ids: list[int] | None = None,
held_rule_ids: list[int] | None = None,
seen_ruling_systems: list[int] | None = None,
) -> dict:
"""Prior-art hint for the plugin's PreToolUse hook on Write/Edit.
@@ -2253,7 +2255,7 @@ async def build_write_path_hint(
cfg = await get_writepath_config(user_id)
empty = {"context": "", "note_ids": [], "sync_note_ids": [], "config": cfg,
"suggested": [], "divergence": [], "derive": [], "derive_keys": [],
"rule_ids": []}
"rule_ids": [], "ruling_system_ids": []}
path = (path or "").strip()
if not cfg["enabled"] or not path:
return empty
@@ -2590,9 +2592,21 @@ async def build_write_path_hint(
design_text, design_dedup = await _design_arm(
user_id, project_id, path, set(exclude_derive or []),
)
# The rulings arm (milestone 444), decided here for the design arm's
# reasons: a lookup by path, so a write that matched no prior art still
# carries its area's rulings, and it never switches the ranked arms on.
rulings = await system_rulings_svc.rulings_for_paths(
user_id, project_id, [path], seen=seen_ruling_systems,
source="rulings_write_path",
)
if not staleness and not synced and not menu and not suggested and not divergence and not derive:
if design_text:
return {**empty, "context": design_text, "derive_keys": [design_dedup]}
if design_text or rulings["lines"]:
return {
**empty,
"context": "\n".join(rulings["lines"] + ([design_text] if design_text else [])),
"derive_keys": [design_dedup] if design_dedup else [],
"ruling_system_ids": rulings["system_ids"],
}
return empty
owners = await owner_names_for({
@@ -2614,7 +2628,9 @@ async def build_write_path_hint(
# Seeded with the staleness line, which is decided above the early
# return and so cannot wait for this list to exist.
lines: list[str] = list(staleness)
# First after staleness: it BINDS, where everything below is prior art.
# First after staleness: rulings and the design system BIND, where
# everything below is prior art.
lines.extend(rulings["lines"])
if design_text:
lines.append(design_text)
sync_note_ids: list[int] = []
@@ -2890,6 +2906,7 @@ async def build_write_path_hint(
),
"rule_ids": rule_ids,
"checkpoint": checkpoint,
"ruling_system_ids": rulings["system_ids"],
}
@@ -2901,18 +2918,46 @@ async def build_tool_rule_hint(
project_id: int = 0,
exclude_rule_ids: list[int] | None = None,
held_rule_ids: list[int] | None = None,
root: str = "",
cwd: str = "",
seen_ruling_systems: list[int] | None = None,
) -> dict:
"""Standing rules that may apply to the action about to be taken — matched
directly (`_tool_rule_hint`, where the design is written), then reached
through a linked lesson (`_rules_via_lessons`, #4633)."""
through a linked lesson (`_rules_via_lessons`, #4633) — and the rulings of
any area whose files the command names (milestone 444).
`root` and `cwd` are the repo's absolute root and the command's working
directory, from the hook; they are what turn the paths a command names
into the repo-relative paths a System's patterns are written in."""
out = await _tool_rule_hint(
user_id, tool_name, command, project_id=project_id,
exclude_rule_ids=exclude_rule_ids, held_rule_ids=held_rule_ids,
)
return await _add_rules_via_lessons(
out = await _add_rules_via_lessons(
user_id, out, project_id=project_id, exclude_rule_ids=exclude_rule_ids,
held_rule_ids=held_rule_ids, where=f"to this {tool_name} call",
)
# Its own arm, not part of the rule search: a lookup by path, with no
# floor and no budget, so it adds to the rule lines rather than competing
# with them. First, because a ruling is the operator's own decision.
# `ruling_system_ids` is set only when a ruling was shown: this arm fires
# on every command, and the hook reads an absent list as an empty one.
try:
paths = system_rulings_svc.command_paths(command, root=root, cwd=cwd) if project_id else []
if paths and (await get_writepath_config(user_id)).get("enabled"):
rulings = await system_rulings_svc.rulings_for_paths(
user_id, project_id, paths,
seen=seen_ruling_systems, source="rulings_pre_tool",
)
if rulings["lines"]:
out["context"] = "\n".join(
rulings["lines"] + ([out["context"]] if out.get("context") else [])
)
out["ruling_system_ids"] = rulings["system_ids"]
except Exception:
logger.debug("pre-tool rulings arm failed", exc_info=True)
return out
async def _tool_rule_hint(
+196
View File
@@ -0,0 +1,196 @@
"""An area's rulings, delivered when its files are touched (milestone 444).
A RULING is the operator's decision about how one area of the work must behave.
It lives in a `Rulings` section at the end of the area's System description
(writing-records.md says how one is written). The description already rides
every record filed under the System; this module is the other half: the first
command or edit in a session that touches the System's files
(`System.path_patterns`) shows its rulings, one line per System.
A LOOKUP, NOT A SEARCH. A shell command scores against prose decisions at noise
level, so nothing here is ranked: a path either falls under a System's
patterns or it does not. It takes no budget and no floor from the retrieval
arms beside it, and writes no retrieval_logs row — there is no score
distribution for it to join. Its surfacings go to `system_usage_events`.
ONCE IN FULL, THEN A REFERENCE (the #3750 shape). The hook keeps the Systems
already shown this session and passes them back; a repeat renders as one short
line rather than the rulings again, and is not counted as a surfacing.
Fails open everywhere: a delivery aid must never break the act it rides on.
"""
from __future__ import annotations
import logging
import posixpath
import re
import shlex
from scribe.services import systems as systems_svc
from scribe.services.system_usage import record_system_surfaced
from scribe.services.text import elide
logger = logging.getLogger(__name__)
# The section heading: `Rulings`, optionally as a markdown heading, bold, or
# with a trailing colon. The LAST one wins — the section sits at the end.
_HEADING = re.compile(r"^[ \t]*(?:#{1,6}[ \t]*)?\**Rulings\**[ \t]*:?[ \t]*$", re.I | re.M)
_BULLET = re.compile(r"^[ \t]*[-*][ \t]+(.*)$")
# Per line, so one System with a long list cannot crowd the others out.
RULING_CHARS = 300
RULINGS_PER_SYSTEM = 8
# A command is not a file list: these bound how much of it is read as paths.
_MAX_PATHS = 40
_LINE_SUFFIX = re.compile(r":\d+(?::\d+)?$")
_HAS_EXTENSION = re.compile(r"\.[A-Za-z0-9]{1,10}$")
_ASSIGNMENT = re.compile(r"^[A-Za-z_][A-Za-z0-9_]*=")
def parse_rulings(description: str | None) -> list[str]:
"""The rulings in a System description, one string per ruling.
A bullet under the last `Rulings` heading is a ruling; an indented line
after one continues it. The section ends at the first line that is
neither — a heading or paragraph written after it is not a ruling.
"""
text = description or ""
found = list(_HEADING.finditer(text))
if not found:
return []
out: list[str] = []
for line in text[found[-1].end():].splitlines():
if not line.strip():
continue
bullet = _BULLET.match(line)
if bullet:
if bullet.group(1).strip():
out.append(bullet.group(1).strip())
elif out and line[:1] in (" ", "\t"):
out[-1] = f"{out[-1]} {line.strip()}"
else:
break
return out
def _relative(token: str, root: str, cwd_rel: str) -> str:
"""`token` as a repo-relative path, or "" when it is not one."""
if token.startswith("/"):
if not root or not token.startswith(root.rstrip("/") + "/"):
return ""
token = token[len(root.rstrip("/")) + 1:]
elif cwd_rel:
token = f"{cwd_rel}/{token}"
path = posixpath.normpath(token)
if path in (".", "") or path == ".." or path.startswith("../"):
return ""
return path
def command_paths(command: str, *, root: str = "", cwd: str = "") -> list[str]:
"""The repo-relative paths a shell command names, in order, deduplicated.
Reads and writes alike: a session forms its picture of an area by reading
it, which is exactly when that area's rulings should reach it. A token is
taken as a path when it has a `/` or a file extension; flags, URLs and
globs are not paths. `root` is the repo's absolute root and `cwd` the
command's working directory, both from the hook — an absolute path
outside the root, or a relative one that climbs out of it, is dropped.
"""
command = command or ""
try:
tokens = shlex.split(command, comments=False, posix=True)
except ValueError:
tokens = command.split()
root = (root or "").rstrip("/")
cwd_rel = ""
if root and cwd:
cwd = cwd.rstrip("/")
if cwd.startswith(root + "/"):
cwd_rel = cwd[len(root) + 1:]
out: list[str] = []
for raw in tokens:
token = raw.strip().strip("'\"").rstrip(";,)|&")
if token.startswith("-"):
if "=" not in token:
continue
token = token.split("=", 1)[1]
elif _ASSIGNMENT.match(token):
token = token.split("=", 1)[1]
if not token or "://" in token or any(c in token for c in "*?[]{}$`<>"):
continue
# `file.py:120` and a test id `file.py::test_x` name the file.
token = _LINE_SUFFIX.sub("", token.split("::", 1)[0])
if "/" not in token and not _HAS_EXTENSION.search(token):
continue
path = _relative(token, root, cwd_rel)
if path and path not in out:
out.append(path)
if len(out) >= _MAX_PATHS:
break
return out
def _full_line(system, path: str, rulings: list[str]) -> str:
shown = [elide(r, RULING_CHARS)[0] for r in rulings[:RULINGS_PER_SYSTEM]]
more = len(rulings) - len(shown)
listed = " ".join(f"({i}) {r}" for i, r in enumerate(shown, 1))
if more > 0:
listed += f" (+{more} more in `get_system({system.id})`)"
return (
f"> Rulings for `{path}` — the operator's decisions about "
f"{system.name} (System {system.id}): {listed} "
"Work here keeps to them; something that would depart from one is a "
"question for the operator, not an option to choose. "
"(Shown once per session.)"
)
def _reference_line(system, path: str) -> str:
return (
f"> `{path}` is in {system.name} (System {system.id}) — its rulings "
"were shown earlier this session and still apply."
)
async def rulings_for_paths(
user_id: int,
project_id: int,
paths: list[str],
*,
seen: list[int] | set[int] | None = None,
source: str,
) -> dict:
"""Rulings lines for the Systems whose files `paths` touch.
Returns {"lines": [...], "system_ids": [...]} — `system_ids` are the
Systems shown IN FULL on this call, for the hook to add to the session's
seen list and for the usage ledger; a System in `seen` gets a reference
line and is in neither. A System with patterns but no rulings says
nothing: there is no decision to deliver, and the charter already rides
its records.
"""
out: dict = {"lines": [], "system_ids": []}
if not project_id or not paths:
return out
try:
already = {int(s) for s in (seen or [])}
for system, hit in await systems_svc.systems_for_paths(user_id, project_id, paths):
rulings = parse_rulings(system.description)
if not rulings:
continue
if system.id in already:
out["lines"].append(_reference_line(system, hit[0]))
else:
out["lines"].append(_full_line(system, hit[0], rulings))
out["system_ids"].append(system.id)
if out["system_ids"]:
record_system_surfaced(
user_id=user_id, system_ids=out["system_ids"], source=source,
project_id=project_id,
)
except Exception:
logger.debug("rulings arm failed", exc_info=True)
return {"lines": [], "system_ids": []}
return out
+67
View File
@@ -0,0 +1,67 @@
"""System usage telemetry — were an area's rulings read once shown?
The twin of `note_usage` and `rule_usage` for Systems (milestone 444). The
rulings arm shows a System's rulings when a command or edit touches the
System's files; a pull is somebody then opening the System (`get_system`). The
ratio says whether delivering rulings by path earns its line.
Fire-and-forget like its siblings: telemetry never adds latency to, or
breaks, the surface it observes, and failures report through the shared
canary rather than vanishing.
"""
from __future__ import annotations
import logging
from scribe.models import async_session
from scribe.models.system_usage import PULLED, SURFACED, SystemUsageEvent
from scribe.services.background import report_telemetry_failure, spawn
logger = logging.getLogger(__name__)
async def _insert_events(rows: list[dict]) -> None:
"""Persist usage rows. Best-effort: failures degrade, visibly."""
try:
async with async_session() as session:
session.add_all([SystemUsageEvent(**row) for row in rows])
await session.commit()
except Exception:
await report_telemetry_failure("system_usage", "write")
def _rows(user_id, system_ids, event: str, source: str, project_id) -> list[dict]:
pid = int(project_id or 0) or None
return [
{"user_id": user_id, "system_id": int(sid), "event": event,
"source": source, "project_id": pid}
for sid in system_ids
]
def record_system_surfaced(
*, user_id: int | None, system_ids: list[int], source: str,
project_id: int | None = None,
) -> None:
"""Fire-and-forget: these Systems' rulings were shown in full. A repeat
rendered as a short reference is not a surfacing and is not recorded."""
try:
rows = _rows(user_id, system_ids, SURFACED, source, project_id)
except Exception:
logger.debug("system usage payload build failed", exc_info=True)
return
if rows:
spawn(_insert_events(rows), site="system_usage_write")
def record_system_pulled(
*, user_id: int | None, system_id: int, source: str,
project_id: int | None = None,
) -> None:
"""Fire-and-forget: a System was opened in full."""
try:
rows = _rows(user_id, [system_id], PULLED, source, project_id)
except Exception:
logger.debug("system usage payload build failed", exc_info=True)
return
spawn(_insert_events(rows), site="system_usage_write")