feat(ledger): write-path stamping — a pulled canon the session then instantiates lands as a hook instance row (#2791, milestone 294 step 5)
CI & Build / Plugin hooks (push) Failing after 2s
CI & Build / Python lint (push) Successful in 4s
CI & Build / integration (push) Failing after 28s
CI & Build / TypeScript typecheck (push) Successful in 36s
CI & Build / Python tests (push) Successful in 1m5s
CI & Build / Build & push image (push) Successful in 39s

The prior-art hook now names the shapes being written (shapes=kind:name —
every definition in the payload, or the one enclosing an Edit found by
walking the file upward) and the server stamps them as instance rows when
the session PULLED a snippet inside PULL_WINDOW that the payload references
by symbol or that the semantic arm scored for this very payload.
classified_by=hook, evidence in reason; never overrides a judgment or a
canonical row, overridable by classify_shapes. Offered-but-unopened stamps
nothing. Pulled-and-already-seen snippets stay in the semantic query as
evidence without re-entering the deduped menu. A brand-new shape gets a
provisional row the next sync confirms or vanishes. Read-scoped keys get
the hint, never the stamp. Plugin 0.1.34.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
2026-08-20 20:19:45 -04:00
co-authored by Claude Fable 5
parent 3e21978a9c
commit 475f0857c9
11 changed files with 807 additions and 18 deletions
+1 -1
View File
@@ -1,7 +1,7 @@
{
"name": "scribe",
"description": "Scribe system-of-record for Claude Code: MCP tools over your notes/tasks/projects/rules, a session-start push channel that surfaces your always-on rules + active-project context, process-skills (writing-plans, systematic-debugging, verification, brainstorming, reusing-code), and your saved Scribe Processes auto-surfaced as skills (/scribe:sync). Replaces superpowers + file-memory with one app-backed plugin.",
"version": "0.1.33",
"version": "0.1.34",
"author": { "name": "Bryan Van Deusen" },
"mcpServers": {
"scribe": {
+56 -6
View File
@@ -14,6 +14,12 @@
# on an instance with no forge connection (decision #2707). Everything else is
# the REUSE menu. The two dedup separately (see the state files below).
#
# It is also the shape ledger's write-path feed (#2791): it names the
# definitions being written (`shapes=`), and the server — only when the
# session has PULLED a snippet this code references or resembles — records
# them as instance rows, classified_by=hook. Evidence, not judgment; the
# context line says what landed so a wrong stamp is corrected in the moment.
#
# NEVER BLOCKS. It returns `additionalContext` with no `permissionDecision`, so
# the write proceeds untouched and Claude sees the note beside the tool result.
# Any failure — unconfigured, unreachable, malformed — exits 0 in silence. A
@@ -96,10 +102,14 @@ fi
# `ReturnType name(...)`) needs a real parser, and `impl` blocks are excluded
# because several per type is normal Rust, not duplication.
# ---------------------------------------------------------------------------
local_lines=""
if [ -n "$repo_root" ] && [ -n "$code" ]; then
# kind<TAB>name for each thing this payload DEFINES.
names=$(printf '%s' "$code" | awk '
# kind<TAB>name for each thing a piece of code DEFINES, in source order. One
# program, two consumers: the local duplicate arm (every definition in the
# payload) and the ledger feed (#2791, below: the definitions being written,
# or the one enclosing an Edit). Rule-for-rule mirrored by the server's
# services/coverage.py extract_shapes — ledger rows are keyed by what THAT
# sees, so the two must agree on what counts as a definition.
scribe_defs() {
awk '
{
# CSS class definition: .name { or .name,
if (match($0, /^[[:space:]]*\.[A-Za-z][A-Za-z0-9_-]*[[:space:]]*[,{]/)) {
@@ -132,8 +142,16 @@ if [ -n "$repo_root" ] && [ -n "$code" ]; then
if (t != "") print "sym\t" t; next
}
}
' 2>/dev/null | sort -u | head -12) || names=""
' 2>/dev/null
}
names=""
if [ -n "$code" ]; then
names=$(printf '%s' "$code" | scribe_defs | sort -u | head -12) || names=""
fi
local_lines=""
if [ -n "$repo_root" ] && [ -n "$names" ]; then
while IFS=$'\t' read -r kind name; do
[ -n "${name:-}" ] || continue
case "$kind" in
@@ -156,6 +174,38 @@ if [ -n "$local_lines" ]; then
local_context="> Already defined elsewhere in this repo — check before adding another copy (\`git grep\` shown; this is a nudge, not a gate):"$'\n'"${local_lines}"
fi
# ---------------------------------------------------------------------------
# THE LEDGER FEED (#2791). The server keeps a shape ledger — every definition
# in the bound repo, classified against recorded canon — and this hook is the
# one place that sees a shape AT THE MOMENT IT IS WRITTEN. So it names the
# shapes in play: every definition in the payload, or — for an Edit that
# changes the inside of a function rather than its signature — the definition
# enclosing the edit, found by walking the target file upward from the edited
# lines. The server decides whether evidence exists (the session pulled a
# snippet this code references or resembles) and stamps instance rows; with
# no pulled canon in play, nothing is recorded. Titles only still — this sends
# names, not bodies.
# ---------------------------------------------------------------------------
shapes="$names"
if [ -z "$shapes" ] && [ -f "$file_path" ] && command -v tac >/dev/null 2>&1; then
old_first=$(printf '%s' "$event" \
| jq -r '.tool_input.old_string // .tool_input.old_str // empty' 2>/dev/null \
| grep -m1 -v '^[[:space:]]*$') || old_first=""
if [ -n "$old_first" ]; then
ln=$(grep -nF -m1 -- "$old_first" "$file_path" 2>/dev/null | cut -d: -f1) || ln=""
if [ -n "$ln" ]; then
shapes=$(head -n "$ln" "$file_path" | tac | scribe_defs | head -1) || shapes=""
fi
fi
fi
shapes_q=""
if [ -n "$shapes" ]; then
enc=$(printf '%s\n' "$shapes" \
| awk -F'\t' 'NF>=2 {printf "%s%s:%s", (n++?",":""), $1, $2}' \
| jq -sRr '@uri' 2>/dev/null) || enc=""
[ -n "$enc" ] && shapes_q="&shapes=${enc}"
fi
url=${SCRIBE_URL:-${CLAUDE_PLUGIN_OPTION_API_ENDPOINT:-}}
token=${SCRIBE_TOKEN:-${CLAUDE_PLUGIN_OPTION_API_TOKEN:-}}
# Guard against an unexpanded ${...} placeholder arriving as a literal.
@@ -230,7 +280,7 @@ fi
# finding that needed no instance to produce.
body=$(curl -fsS --max-time 5 \
-H "Authorization: Bearer ${token}" \
"${url%/}/api/plugin/prior-art?path=${path_enc}&code=${code_enc}${repo_q}${exclude_q}${sync_exclude_q}" 2>/dev/null) || body=""
"${url%/}/api/plugin/prior-art?path=${path_enc}&code=${code_enc}${repo_q}${exclude_q}${sync_exclude_q}${shapes_q}" 2>/dev/null) || body=""
context=""
if [ -n "$body" ]; then
+3
View File
@@ -34,6 +34,9 @@ through recall/auto-inject; this skill is the active reflex around that.
to ask both at once.
- If a snippet fits, pull it in full with `get_snippet(id)` and reuse it — its
`location` points at the reference implementation. Adapt, don't re-derive.
The pull also does the accounting: the code you then write that references
or resembles it is stamped an `instance` of that canon in the shape ledger
(classified_by=hook) — reuse from memory leaves no row.
- If auto-inject already surfaced a snippet title that looks relevant, that's
your cue to `get_snippet` it rather than start from scratch.
- **Prior art offered beside a write is not noise — read it.** When Scribe notes
+17
View File
@@ -34,6 +34,23 @@ row carries a status:
nothing. Rows, never prose — a consumer list in a note or verification
detail cannot be sorted, queried, or diffed.
## Rows that arrive on their own
Two feeds keep the ledger current between your batches, so most shapes never
need a hand judgment:
- **The sync** stamps a snippet's own reference location `canonical`
(`classified_by: mechanical`).
- **The write path** stamps instances as you work: when you `get_snippet` a
canon and then Write/Edit code that references or resembles it, the
definitions being written land as `instance` rows (`classified_by: hook`,
the evidence in `reason`), and the prior-art hook tells you what landed
("Shape accounting: recorded at … → instance of #N"). Offered-but-unopened
snippets stamp nothing — so *pull the canon you are instantiating*; that
pull is what turns your reuse into accounting. A hook row is evidence, not
judgment: it never overrides a classification you made, and a
`classify_shapes` call overrides it.
## The derive-first rule
N same-shaped occurrences matching **no** recorded canon is never N loose
+11 -4
View File
@@ -77,10 +77,17 @@ async def list_shapes(
limit/offset: page through big ledgers (limit caps at 500).
Returns {"shapes": [...], "total": N} — total counts every match, not
just this page. Classify what you can judge with classify_shapes; a
repeating shape with NO recorded canon is a derive-one-first moment
(consolidate onto a reference, create_snippet it, then classify the
rest against it), never N loose classifications.
just this page. Each row's `classified_by` says who judged: agent /
audit / import are judgments; `mechanical` is the canonical stamp the
sync applies; `hook` is write-path EVIDENCE (#2791) — the session pulled
a snippet and then wrote code referencing/resembling it, so the shape
was stamped an instance with the evidence in `reason`. A hook row is
overridable by any classify_shapes call; it never overrides yours.
Classify what you can judge with classify_shapes; a repeating shape
with NO recorded canon is a derive-one-first moment (consolidate onto a
reference, create_snippet it, then classify the rest against it), never
N loose classifications.
"""
uid = current_user_id()
rows, total = await shape_ledger_svc.list_project_shapes(
+35
View File
@@ -127,6 +127,15 @@ async def write_path_prior_art():
surfaced. A separate channel on purpose: a reuse
hint shown early must not suppress the record-sync
nudge when the recorded file is edited later.
shapes (opt) — comma-separated `kind:name` definitions the hook
found in (or enclosing) the payload, kind being
css|sym. The shape ledger's write-path feed
(#2791): when the session recently PULLED a
snippet this payload references or resembles,
these land as instance rows (classified_by=hook).
Honoured only for a caller allowed to write — a
read-scoped key still gets the hint, and never
changes accounting on a GET.
"""
path = (request.args.get("path") or "").strip()
code = request.args.get("code") or ""
@@ -149,14 +158,40 @@ async def write_path_prior_art():
int(p) for p in (request.args.get("exclude_sync_ids") or "").split(",")
if p.strip().isdigit()
]
shapes = _parse_shapes(request.args.get("shapes") or "")
api_key = getattr(g, "api_key", None)
may_stamp = api_key is None or getattr(api_key, "scope", "") == "write"
result = await plugin_ctx_svc.build_write_path_hint(
g.user.id, path, code=code, project_id=project_id,
exclude_ids=exclude_ids, exclude_sync_ids=exclude_sync_ids,
stamp_shapes=shapes if may_stamp else None,
repo_key=repo_bindings_svc.normalize_repo_key(repo) if repo else "",
)
return jsonify(result)
# The hook names at most a dozen definitions per write; anything past that is
# a generated file, not a shape being instantiated.
_SHAPES_CAP = 12
def _parse_shapes(raw: str) -> list[tuple[str, str]]:
"""`css:btn-primary,sym:onTrash` → [("css", "btn-primary"), ("sym", "onTrash")].
Unknown kinds and empty names are dropped, duplicates collapse, and the
list is capped — the hook's own cap, re-applied so the contract holds
for any caller."""
out: list[tuple[str, str]] = []
for part in raw.split(","):
kind, _sep, name = part.strip().partition(":")
kind, name = kind.strip(), name.strip()
if kind in ("css", "sym") and name and (kind, name) not in out:
out.append((kind, name))
if len(out) >= _SHAPES_CAP:
break
return out
@plugin_bp.get("/processes")
@login_required
async def process_manifest():
+71 -4
View File
@@ -27,6 +27,7 @@ from scribe.services import knowledge as knowledge_svc
from scribe.services import notes as notes_svc
from scribe.services import projects as projects_svc
from scribe.services import rulebooks as rulebooks_svc
from scribe.services import shape_ledger as shape_ledger_svc
from scribe.services import snippets as snippets_svc
from scribe.services.access import label_shared_items, owner_names_for
from scribe.services.embeddings import semantic_search_notes
@@ -703,6 +704,8 @@ async def build_write_path_hint(
project_id: int = 0,
exclude_ids: list[int] | None = None,
exclude_sync_ids: list[int] | None = None,
stamp_shapes: list[tuple[str, str]] | None = None,
repo_key: str = "",
) -> dict:
"""Prior-art hint for the plugin's PreToolUse hook on Write/Edit.
@@ -749,9 +752,20 @@ async def build_write_path_hint(
un-scored surfacing now has its own home: every arm emits note_usage_events,
tagged 'write_path_sync' vs 'write_path_place' vs 'write_path_semantic', so
each claim's pull-through rate is measurable on its own.
``stamp_shapes`` turns the same request into the ledger's write-path feed
(#2791): the (kind, name) definitions the hook saw in — or enclosing —
the payload. When the session has PULLED a snippet recently and this
payload references or resembles it, those shapes land as `instance` rows
(classified_by=hook, see shape_ledger.stamp_write_path_instances) and
the result's ``stamped`` lists them. The route passes it only for a
caller allowed to write — a read-scoped key gets the hint, never the
stamp. ``repo_key`` (the hook's remote, normalised) homes a provisional
row for a shape the ledger has not synced yet.
"""
cfg = await get_writepath_config(user_id)
empty = {"context": "", "note_ids": [], "sync_note_ids": [], "config": cfg}
empty = {"context": "", "note_ids": [], "sync_note_ids": [], "config": cfg,
"stamped": []}
path = (path or "").strip()
if not cfg["enabled"] or not path:
return empty
@@ -800,6 +814,15 @@ async def build_write_path_hint(
seen.add(nid)
placed.append(("nearby", item))
# The stamping feed's "actually pulled it" half (#2791). Read once, before
# the semantic arm, because the arm's query doubles as the resemblance
# test: a pulled snippet this session already saw (so it sits in `seen`)
# must still be SCORED for this payload — it just isn't re-listed.
pulled: dict = {}
if stamp_shapes:
pulled = await shape_ledger_svc.recent_pulls(user_id)
resembles: dict[int, float] = {}
# --- arm 2: by meaning ---
scored: list[tuple[str, dict]] = []
remaining = top_k - len(synced) - len(placed)
@@ -821,12 +844,15 @@ async def build_write_path_hint(
query = concept_query(query) or query
if remaining > 0 and query:
t0 = time.perf_counter()
# Pulled-and-seen ids stay in the query (as evidence) but never in
# the menu — the dedup contract holds, the resemblance still lands.
pulled_seen = seen & set(pulled)
hits = await semantic_search_notes(
user_id, query,
limit=remaining,
limit=remaining + len(pulled_seen),
threshold=cfg["threshold"],
project_id=scope_project,
exclude_ids=seen,
exclude_ids=seen - pulled_seen,
# Snippets AND recorded experience (#2246). This arm was
# snippets-only, which is auto-inject's mistake inverted: an issue
# saying "we tried this and it deadlocked", or a dev-log recording
@@ -845,6 +871,11 @@ async def build_write_path_hint(
# the browse scope and never surfaces a one-to-one direct share.
scope="browse",
)
resembles = {
int(note.id): float(score) for score, note in hits
if int(note.id) in pulled
}
hits = [(s, n) for s, n in hits if int(n.id) not in seen][:remaining]
record_retrieval(
user_id=user_id, source="write_path", query=query,
threshold=cfg["threshold"], limit=remaining,
@@ -879,7 +910,20 @@ async def build_write_path_hint(
))
menu = (placed + scored)[:max(0, top_k - len(synced))]
if not synced and not menu:
# The stamp runs whether or not anything is rendered — after dedup, the
# common case is a silent hint and a pulled canon being instantiated.
stamped: list[dict] = []
if stamp_shapes and pulled:
try:
stamped = await shape_ledger_svc.stamp_write_path_instances(
user_id, project_id, path=path, shapes=stamp_shapes,
code=code or "", pulled=pulled, resembles=resembles,
repo_key=repo_key,
)
except Exception:
logger.warning("Write-path ledger stamping failed", exc_info=True)
if not synced and not menu and not stamped:
return empty
owners = await owner_names_for({
@@ -943,6 +987,9 @@ async def build_write_path_hint(
note_ids.append(int(item["id"]))
lines.append(_prior_art_line(item, marker, owner, foreign_lang))
if stamped:
lines.append(_stamp_line(path, stamped))
# Split by arm, which is the whole reason this table exists. The place arm
# carries no score and so has no home in retrieval_logs; before #2085 a
# snippet surfaced BY PLACE left no trace anywhere, making the arm that
@@ -964,9 +1011,29 @@ async def build_write_path_hint(
"note_ids": note_ids,
"sync_note_ids": sync_note_ids,
"config": cfg,
"stamped": stamped,
}
def _stamp_line(path: str, stamped: list[dict]) -> str:
"""One line saying what the ledger just recorded, so the session can
correct a wrong stamp in the moment rather than an audit finding it."""
by_snippet: dict[int, list[str]] = {}
for row in stamped:
label = f".{row['symbol']}" if row["kind"] == "css" else row["symbol"]
by_snippet.setdefault(int(row["snippet_id"]), []).append(f"`{label}`")
parts = [
f"{', '.join(names)} → instance of #{sid}"
for sid, names in by_snippet.items()
]
return (
f"> Shape accounting: recorded at `{path}` — {'; '.join(parts)} "
"(classified_by=hook: you pulled that snippet this session and this "
"code references/resembles it). Not an instance? `classify_shapes` "
"overrides a hook stamp."
)
async def _topic_titles(topic_ids: set[int]) -> dict[int, str]:
"""Map topic_id -> title for the given ids (live topics only)."""
if not topic_ids:
+189 -1
View File
@@ -21,13 +21,17 @@ an extracted shape.
"""
from __future__ import annotations
from datetime import datetime, timezone
import logging
import re
from datetime import datetime, timedelta, timezone
from sqlalchemy import select
from scribe.models import async_session
from scribe.models.code_shape import CodeShape
logger = logging.getLogger(__name__)
# Statuses whose meaning requires a snippet target.
_NEEDS_TARGET = ("canonical", "instance", "variant")
@@ -408,3 +412,187 @@ async def snippet_consumers(user_id: int, note_id: int) -> dict:
_consumer_dict(row)
)
return out
# --- write-path stamping (#2791): hook evidence lands as rows ----------------
#
# The write-path hook (plugin/hooks/scribe_prior_art.sh) fires on every
# Write/Edit and already carries the two halves of a consumer-map row: the
# file being written and the definitions in (or enclosing) the payload. What
# it could not say on its own is WHICH canon the session is instantiating.
# The pull stream answers that: a snippet the session opened in full
# (get_snippet) and is now writing code that references or resembles is being
# reused — and a reused canon's call site is an `instance` (note 2786).
#
# The rule, deliberately two-sided so it cannot fire on noise:
# PULLED — a PULLED usage event by this user inside PULL_WINDOW. Offered-
# but-ignored (surfaced, never opened) stamps nothing.
# IN PLAY — the payload references the snippet's symbol by name, or the
# semantic arm scored it above the write-path threshold for this
# very payload. Either is evidence; the pull alone is not.
# Both hold → every shape the hook named at that path, of the snippet's kind,
# becomes instance-of-N with classified_by="hook" and the evidence as reason.
#
# A hook row is EVIDENCE, not judgment: it only ever lands on rows nobody has
# judged (unclassified) or rows an earlier hook stamped, never on a canonical
# row or an agent/audit/import judgment. Re-judge with classify_shapes.
# "The write path actually pulled it": a working session's reach. The
# precision comes from the in-play test above, not from this window.
PULL_WINDOW = timedelta(hours=6)
def snippet_kind(symbol: str, language: str) -> str:
"""The ledger kind a snippet's reference belongs to — "css" when its
symbol is a class selector (or it is a stylesheet with no symbol),
else "sym"."""
sym = (symbol or "").strip()
if sym.startswith("."):
return "css"
if not sym and (language or "").strip().lower() in ("css", "scss", "sass", "less"):
return "css"
return "sym"
def references_symbol(code: str, symbol: str, kind: str) -> bool:
"""Does this payload name the snippet's symbol? Word-bounded so `confirm`
never claims `confirmed`; a CSS class matches as `.btn` or inside a class
attribute (`btn btn-primary`), dashes counting as part of the name."""
sym = _norm_symbol(symbol or "")
if not sym or not code:
return False
if kind == "css":
pattern = rf"(?<![\w-]){re.escape(sym)}(?![\w-])"
else:
pattern = rf"(?<![\w$]){re.escape(sym)}(?![\w$])"
return re.search(pattern, code) is not None
async def recent_pulls(user_id: int, *, window: timedelta = PULL_WINDOW) -> dict[int, datetime]:
"""{note_id: last pulled at} for every note this user opened in full
inside ``window`` — the "actually pulled it" half of the stamping rule.
Reads the usage telemetry table; an unreadable table means no evidence."""
from sqlalchemy import func
from scribe.models.note_usage import PULLED, NoteUsageEvent
since = datetime.now(timezone.utc) - window
try:
async with async_session() as session:
rows = await session.execute(
select(NoteUsageEvent.note_id, func.max(NoteUsageEvent.created_at))
.where(
NoteUsageEvent.user_id == user_id,
NoteUsageEvent.event == PULLED,
NoteUsageEvent.created_at >= since,
)
.group_by(NoteUsageEvent.note_id)
)
return {int(nid): ts for nid, ts in rows.all()}
except Exception:
logger.warning("recent_pulls read failed — no hook stamping this write", exc_info=True)
return {}
async def stamp_write_path_instances(
user_id: int,
project_id: int,
*,
path: str,
shapes: list[tuple[str, str]],
code: str,
pulled: dict[int, datetime],
resembles: dict[int, float] | None = None,
repo_key: str = "",
) -> list[dict]:
"""Land hook evidence as `instance` rows for the shapes being written.
``shapes`` is the hook's (kind, name) list for ``path``; ``pulled`` is
recent_pulls(); ``resembles`` maps snippet ids the semantic arm scored
for this payload to their score. Returns the rows stamped, each
{path, symbol, kind, snippet_id, reason} — empty in the common case.
A shape the ledger has no live row for yet (it is being written right
now) gets a PROVISIONAL row under ``repo_key`` — first/last-seen unset —
so the stamp is not lost to the next sync, which either confirms the
shape (sets its seen marker) or stamps it vanished. No repo key → only
existing rows are stamped.
When more than one pulled snippet is in play for a shape, a by-name
reference beats resemblance and the most recent pull breaks ties: a row
holds one canon (the known model limit logged on #2790).
"""
from scribe.services import access
from scribe.services import snippets as snippets_svc
from scribe.services.snippets import snippet_fields
resembles = resembles or {}
path = (path or "").strip()
wanted = [(k, n.strip()) for k, n in shapes if k in ("css", "sym") and n.strip()]
if not project_id or not path or not wanted or not pulled:
return []
if not await access.can_write_project(user_id, project_id):
return []
# Which pulled canons are in play for this payload, by kind, ranked.
in_play: dict[str, list[tuple[int, datetime, int, str]]] = {}
for sid, pulled_at in pulled.items():
note = await snippets_svc.get_snippet(user_id, sid)
if note is None:
continue
fields = snippet_fields(note)
symbol = fields.get("symbol") or ""
kind = snippet_kind(symbol, fields.get("language") or "")
if references_symbol(code, symbol, kind):
rank, why = 2, f"hook: pulled #{sid}; payload references `{_norm_symbol(symbol)}`"
elif sid in resembles:
rank, why = 1, f"hook: pulled #{sid}; payload resembles it ({resembles[sid]:.2f})"
else:
continue
in_play.setdefault(kind, []).append((rank, pulled_at, sid, why))
if not in_play:
return []
for bucket in in_play.values():
bucket.sort(key=lambda t: (t[0], t[1]), reverse=True)
now = datetime.now(timezone.utc)
stamped: list[dict] = []
async with async_session() as session:
rows = (
await session.execute(
select(CodeShape).where(
CodeShape.project_id == project_id,
CodeShape.path == path,
CodeShape.vanished_at.is_(None),
)
)
).scalars().all()
by_key = {(r.symbol, r.kind): r for r in rows}
for kind, name in wanted:
bucket = in_play.get(kind)
if not bucket:
continue
_rank, _at, sid, why = bucket[0]
row = by_key.get((name, kind))
if row is None:
if not repo_key:
continue
row = CodeShape(
project_id=project_id, repo_key=repo_key,
path=path, symbol=name, kind=kind,
)
session.add(row)
by_key[(name, kind)] = row
elif not (row.status == "unclassified" or row.classified_by == "hook"):
continue # a judgment — or the canon itself — stands
row.status = "instance"
row.snippet_id = sid
row.reason = why
row.classified_by = "hook"
row.classified_at = now
stamped.append({
"path": path, "symbol": name, "kind": kind,
"snippet_id": sid, "reason": why,
})
if stamped:
await session.commit()
return stamped
+130
View File
@@ -219,3 +219,133 @@ async def test_sync_refiles_rows_whose_snippet_was_purged(seeded):
))).scalar_one()
assert row.status == "unclassified"
assert row.snippet_id is None
# --- #2791: the write-path feed lands hook evidence as rows -------------------
@pytest.mark.integration
async def test_write_path_stamp_is_evidence_that_yields_to_judgment(seeded):
"""Pulled + referenced → every named shape of the snippet's kind becomes
an instance row, classified_by=hook, carrying the evidence as reason. A
later agent judgment on one of them stands against a re-stamp; the hook
may only overwrite nobody's judgment or its own. The outsider stamps
nothing (write-gated like every other ledger write)."""
from datetime import datetime, timezone
from scribe.services.shape_ledger import stamp_write_path_instances
owner, other, pid, sid = (
seeded["owner"], seeded["other"], seeded["pid"], seeded["snippet"]
)
pulled = {sid: datetime.now(timezone.utc)}
code = "app = factory()\nreturn app\n" # references the snippet's symbol
assert await stamp_write_path_instances(
other, pid, path="src/app.py", shapes=[("sym", "make_app")],
code=code, pulled=pulled,
) == []
stamped = await stamp_write_path_instances(
owner, pid, path="src/app.py",
shapes=[("sym", "make_app"), ("sym", "Config"), ("css", "nope")],
code=code, pulled=pulled,
)
assert {s["symbol"] for s in stamped} == {"make_app", "Config"} # css skipped: no css canon
rows, _ = await list_project_shapes(owner, pid, snippet_id=sid)
by_symbol = {r.symbol: r for r in rows}
assert by_symbol["make_app"].status == "instance"
assert by_symbol["make_app"].classified_by == "hook"
assert by_symbol["make_app"].reason == f"hook: pulled #{sid}; payload references `factory`"
# A judgment lands; the next stamp must leave it alone but may re-stamp
# its own earlier row.
await classify_shapes(owner, pid, [
{"path": "src/app.py", "symbol": "make_app", "status": "exempt",
"reason": "the app factory is its own thing"},
])
again = await stamp_write_path_instances(
owner, pid, path="src/app.py",
shapes=[("sym", "make_app"), ("sym", "Config")], code=code, pulled=pulled,
)
assert {s["symbol"] for s in again} == {"Config"}
rows, _ = await list_project_shapes(owner, pid, path="src/app.py")
by_symbol = {r.symbol: r for r in rows}
assert by_symbol["make_app"].status == "exempt"
assert by_symbol["Config"].status == "instance"
# Neither pulled nor in play → nothing, even with shapes named.
assert await stamp_write_path_instances(
owner, pid, path="src/util.py", shapes=[("sym", "helper")],
code="print('unrelated')", pulled=pulled,
) == []
@pytest.mark.integration
async def test_a_brand_new_shape_gets_a_provisional_row_the_sync_settles(seeded):
"""The shape being written right now has no ledger row yet. With the
hook's repo key it gets a provisional one — seen markers unset — so the
stamp survives until the next sync, which confirms it (sets the marker)
or stamps it vanished. Without a repo key only existing rows are touched."""
from datetime import datetime, timezone
from scribe.services.shape_ledger import stamp_write_path_instances
owner, pid, sid = seeded["owner"], seeded["pid"], seeded["snippet"]
pulled = {sid: datetime.now(timezone.utc)}
code = "def build():\n return factory()\n"
assert await stamp_write_path_instances(
owner, pid, path="src/new.py", shapes=[("sym", "build")],
code=code, pulled=pulled, # no repo_key
) == []
stamped = await stamp_write_path_instances(
owner, pid, path="src/new.py", shapes=[("sym", "build")],
code=code, pulled=pulled, repo_key=REPO,
)
assert [s["symbol"] for s in stamped] == ["build"]
async with async_session() as s:
row = (await s.execute(select(CodeShape).where(
CodeShape.project_id == pid, CodeShape.path == "src/new.py",
))).scalar_one()
assert row.status == "instance" and row.classified_by == "hook"
assert row.first_seen_commit is None and row.last_seen_commit is None
# The sync sees the shape in the tree → confirmed, stamp intact.
await sync_repo_shapes(
pid, REPO, SHAPES + [("src/new.py", "sym", "build")], seen_marker="abc123",
)
rows, _ = await list_project_shapes(owner, pid, path="src/new.py")
assert rows[0].status == "instance" and rows[0].last_seen_commit == "abc123"
# The sync no longer sees it → vanished, out of the live accounting.
await sync_repo_shapes(pid, REPO, SHAPES, seen_marker="def456")
rows, _ = await list_project_shapes(owner, pid, path="src/new.py")
assert rows == []
rows, _ = await list_project_shapes(owner, pid, path="src/new.py", include_vanished=True)
assert rows[0].vanished_at is not None
@pytest.mark.integration
async def test_recent_pulls_reads_the_usage_stream(seeded):
"""The "actually pulled it" half is the PULLED usage event, inside the
window; a surfacing alone is not a pull."""
from datetime import datetime, timedelta, timezone
from scribe.models.note_usage import PULLED, SURFACED, NoteUsageEvent
from scribe.services.shape_ledger import recent_pulls
owner, sid = seeded["owner"], seeded["snippet"]
now = datetime.now(timezone.utc)
async with async_session() as s:
s.add_all([
NoteUsageEvent(user_id=owner, note_id=sid, event=PULLED, source="mcp_get_snippet"),
NoteUsageEvent(user_id=owner, note_id=sid + 1000, event=SURFACED, source="auto_inject"),
NoteUsageEvent(user_id=owner, note_id=sid + 2000, event=PULLED,
source="mcp_get_snippet", created_at=now - timedelta(days=2)),
])
await s.commit()
pulls = await recent_pulls(owner)
assert sid in pulls
assert sid + 1000 not in pulls
assert sid + 2000 not in pulls
+53
View File
@@ -82,3 +82,56 @@ def test_classify_and_list_are_mounted_as_mcp_tools():
mcp = build_mcp_server()
for name in ("classify_shapes", "list_shapes", "refresh_pattern_coverage"):
assert mcp._tool_manager.get_tool(name) is not None
# --- step 5: the write-path feed's evidence tests (pure) ---------------------
def test_symbol_reference_is_word_bounded_and_kind_aware():
from scribe.services.shape_ledger import references_symbol as ref
code = "const ok = await confirmed({ title: 'x' });\nif (!ok) return;"
assert ref(code, "confirmed", "sym")
assert not ref(code, "confirm", "sym") # prefix never claims the call
assert not ref("", "confirmed", "sym")
assert not ref(code, "", "sym")
# CSS: the class as a selector or inside a class attribute; dashes are part
# of the name, so `btn` must not claim `btn-primary`.
html = '<button class="btn btn-primary">Go</button>'
assert ref(html, ".btn-primary", "css")
assert ref(html, "btn-primary", "css")
assert ref(".btn-primary { color: red }", ".btn-primary", "css")
assert not ref('<button class="btn-primary">', ".btn", "css")
assert ref('<button class="btn-primary btn">', ".btn", "css")
def test_snippet_kind_reads_the_symbol_then_the_language():
from scribe.services.shape_ledger import snippet_kind
assert snippet_kind(".btn-primary", "css") == "css"
assert snippet_kind(".btn-primary", "") == "css"
assert snippet_kind("confirmed", "typescript") == "sym"
assert snippet_kind("", "scss") == "css" # whole-stylesheet record
assert snippet_kind("", "python") == "sym"
def test_route_shapes_param_parses_capped_and_deduped():
from scribe.routes.plugin import _SHAPES_CAP, _parse_shapes
assert _parse_shapes("css:btn-primary,sym:onTrash") == [
("css", "btn-primary"), ("sym", "onTrash"),
]
assert _parse_shapes(" sym:a , sym:a ,bogus:x,sym:,:,") == [("sym", "a")]
assert _parse_shapes("") == []
many = ",".join(f"sym:f{i}" for i in range(40))
assert len(_parse_shapes(many)) == _SHAPES_CAP
def test_hook_is_a_server_internal_classifier():
"""`hook` is in the status vocabulary but NOT a via a caller may claim —
a classify_shapes call saying via="hook" would launder judgment as
evidence (the reverse of the stamping rule's point)."""
from scribe.services.shape_ledger import _CALLER_VIAS
assert "hook" in SHAPE_CLASSIFIERS
assert "hook" not in _CALLER_VIAS
+241 -2
View File
@@ -778,11 +778,13 @@ def test_route_reads_every_arg_the_hook_sends():
from scribe.routes import plugin as routes
src = inspect.getsource(routes.write_path_prior_art)
for arg in ("path", "code", "repo", "project_id", "exclude_ids", "exclude_sync_ids"):
for arg in ("path", "code", "repo", "project_id", "exclude_ids",
"exclude_sync_ids", "shapes"):
assert f'request.args.get("{arg}"' in src, f"route ignores {arg}"
hook = HOOK.read_text()
for arg in ("path=", "code=", "repo=", "exclude_ids=", "exclude_sync_ids="):
for arg in ("path=", "code=", "repo=", "exclude_ids=", "exclude_sync_ids=",
"shapes="):
assert arg in hook, f"hook never sends {arg}"
@@ -1009,3 +1011,240 @@ def test_local_arm_finds_duplicates_in_every_language_family(
ctx = json.loads(out.stdout)["hookSpecificOutput"]["additionalContext"]
assert "already defined" in ctx
assert "create_snippet" in ctx
# --- #2791: the write-path feed — hook evidence lands as ledger rows ----------
def _ts():
from datetime import datetime, timezone
return datetime.now(timezone.utc)
@pytest.mark.asyncio
async def test_stamping_needs_named_shapes_and_a_recent_pull():
"""Offered-but-ignored stamps nothing: without a PULLED event there is no
evidence, and without the hook naming shapes there is nothing to stamp.
Neither case may even read the pull stream."""
from scribe.services import plugin_context as pc
pulls = AsyncMock(return_value={})
stamp = AsyncMock(return_value=[])
with patch.object(pc, "get_writepath_config", AsyncMock(return_value=_cfg())), \
patch.object(pc.snippets_svc, "list_snippets", AsyncMock(return_value=([], 0))), \
patch.object(pc, "semantic_search_notes", AsyncMock(return_value=[])), \
patch.object(pc, "record_retrieval", MagicMock()), \
patch.object(pc.shape_ledger_svc, "recent_pulls", pulls), \
patch.object(pc.shape_ledger_svc, "stamp_write_path_instances", stamp):
out = await pc.build_write_path_hint(1, "src/x.py", code=REAL_CODE, project_id=4)
assert out["stamped"] == []
pulls.assert_not_awaited() # no shapes → no read
out = await pc.build_write_path_hint(
1, "src/x.py", code=REAL_CODE, project_id=4,
stamp_shapes=[("sym", "debounce")],
)
pulls.assert_awaited_once()
stamp.assert_not_awaited() # shapes, but no pull
assert out["stamped"] == []
@pytest.mark.asyncio
async def test_a_pulled_snippet_already_seen_is_evidence_not_menu():
"""The pulled-then-written flow IS the dedup-excluded flow: the hint offered
#7 earlier (so it sits in exclude_ids), the session pulled it, and now
writes code resembling it. #7 must be scored for this payload — and handed
to the stamp as resemblance — without being re-listed in the menu."""
from scribe.services import plugin_context as pc
search = AsyncMock(return_value=[(0.91, _note(7, "pulled")), (0.80, _note(8, "fresh"))])
stamp = AsyncMock(return_value=[{
"path": "src/x.py", "symbol": "debounce", "kind": "sym",
"snippet_id": 7, "reason": "hook: pulled #7; payload resembles it (0.91)",
}])
with patch.object(pc, "get_writepath_config", AsyncMock(return_value=_cfg(top_k=3))), \
patch.object(pc.snippets_svc, "list_snippets", AsyncMock(return_value=([], 0))), \
patch.object(pc, "semantic_search_notes", search), \
patch.object(pc, "record_retrieval", MagicMock()), \
patch.object(pc, "owner_names_for", AsyncMock(return_value={})), \
patch.object(pc.shape_ledger_svc, "recent_pulls", AsyncMock(return_value={7: _ts()})), \
patch.object(pc.shape_ledger_svc, "stamp_write_path_instances", stamp):
out = await pc.build_write_path_hint(
1, "src/x.py", code=REAL_CODE, project_id=4, exclude_ids=[7],
stamp_shapes=[("sym", "debounce")], repo_key="git.example.com/a/b",
)
# The query kept #7 eligible (and widened the budget by one for it)...
kw = search.call_args.kwargs
assert 7 not in kw["exclude_ids"]
assert kw["limit"] == 4
# ...but the menu still honours the session dedup.
assert out["note_ids"] == [8]
assert "#7" not in "\n".join(
line for line in out["context"].splitlines() if "[similar" in line
)
# The stamp saw the pull and the resemblance score for this payload.
skw = stamp.call_args.kwargs
assert skw["pulled"] == {7: skw["pulled"][7]}
assert skw["resembles"] == {7: 0.91}
assert skw["shapes"] == [("sym", "debounce")]
assert skw["repo_key"] == "git.example.com/a/b"
assert out["stamped"][0]["snippet_id"] == 7
# And the session is told what landed, with the way to correct it.
assert "Shape accounting" in out["context"]
assert "`debounce` → instance of #7" in out["context"]
assert "classify_shapes" in out["context"]
@pytest.mark.asyncio
async def test_a_stamp_renders_even_when_the_hint_is_otherwise_silent():
"""After dedup the common case is an empty hint; the stamp must still run
and still be reported — silence about accounting is how hook rows would
become invisible."""
from scribe.services import plugin_context as pc
stamped = [{"path": "web/b.css", "symbol": "btn-primary", "kind": "css",
"snippet_id": 5, "reason": "hook: pulled #5; payload references `btn-primary`"}]
with patch.object(pc, "get_writepath_config", AsyncMock(return_value=_cfg())), \
patch.object(pc.snippets_svc, "list_snippets", AsyncMock(return_value=([], 0))), \
patch.object(pc, "semantic_search_notes", AsyncMock(return_value=[])), \
patch.object(pc, "record_retrieval", MagicMock()), \
patch.object(pc, "owner_names_for", AsyncMock(return_value={})), \
patch.object(pc.shape_ledger_svc, "recent_pulls", AsyncMock(return_value={5: _ts()})), \
patch.object(pc.shape_ledger_svc, "stamp_write_path_instances",
AsyncMock(return_value=stamped)):
out = await pc.build_write_path_hint(
1, "web/b.css", code=".btn-primary { color: red; }" * 4, project_id=4,
stamp_shapes=[("css", "btn-primary")],
)
assert out["note_ids"] == []
assert out["stamped"] == stamped
assert "`.btn-primary` → instance of #5" in out["context"]
@pytest.mark.asyncio
async def test_a_failing_stamp_does_not_sink_the_hint():
from scribe.services import plugin_context as pc
with patch.object(pc, "get_writepath_config", AsyncMock(return_value=_cfg())), \
patch.object(pc.snippets_svc, "list_snippets",
AsyncMock(return_value=([_snippet_item(12, "records me")], 1))), \
patch.object(pc, "semantic_search_notes", AsyncMock(return_value=[])), \
patch.object(pc, "record_retrieval", MagicMock()), \
patch.object(pc, "owner_names_for", AsyncMock(return_value={})), \
patch.object(pc.shape_ledger_svc, "recent_pulls", AsyncMock(return_value={12: _ts()})), \
patch.object(pc.shape_ledger_svc, "stamp_write_path_instances",
AsyncMock(side_effect=RuntimeError("ledger down"))):
out = await pc.build_write_path_hint(
1, "src/x.py", code=REAL_CODE, project_id=4, stamp_shapes=[("sym", "f")],
)
assert out["sync_note_ids"] == [12]
assert out["stamped"] == []
def test_route_stamps_only_for_a_caller_allowed_to_write():
"""A read-scoped key gets the hint — every plugin hook works on a read key
— but a GET must never change accounting for it. The route passes the
shapes through only when the key is write-scoped (or it's a session)."""
import inspect
from scribe.routes import plugin as routes
src = inspect.getsource(routes.write_path_prior_art)
assert 'request.args.get("shapes"' in src
assert '== "write"' in src
assert "stamp_shapes=shapes if may_stamp else None" in src
assert "normalize_repo_key(repo)" in src
hook = HOOK.read_text()
assert "&shapes=" in hook
def test_hook_names_the_shapes_being_written():
"""The feed's two inputs: every definition in the payload, or — for an Edit
that changes a body, not a signature — the definition enclosing the edit,
found by walking the target file upward from the edited lines."""
src = HOOK.read_text()
assert "scribe_defs()" in src # one extractor, two consumers
assert ".tool_input.old_string" in src # the Edit's anchor
assert "| tac | scribe_defs | head -1" in src # nearest definition above
# The ledger feed sends NAMES, never bodies, and stays on the one GET.
assert src.count("/api/plugin/prior-art?") == 1
def _run_hook_against_sink(tmp_path, payload):
"""Run the hook with SCRIBE_URL pointed at a throwaway local listener and
return the query the hook sent. Lets the shell be tested end to end —
the extraction, the encoding, the URL — without a Scribe instance."""
import http.server
import threading
import urllib.parse
seen: dict = {}
class _Sink(http.server.BaseHTTPRequestHandler):
def do_GET(self):
seen.update(urllib.parse.parse_qs(urllib.parse.urlparse(self.path).query))
self.send_response(200)
self.send_header("Content-Type", "application/json")
self.end_headers()
self.wfile.write(b'{"context":"","note_ids":[]}')
def log_message(self, *a):
pass
server = http.server.HTTPServer(("127.0.0.1", 0), _Sink)
thread = threading.Thread(target=server.serve_forever, daemon=True)
thread.start()
try:
env = dict(_hook_runtime_env(), SCRIBE_URL=f"http://127.0.0.1:{server.server_port}")
out = subprocess.run(
["bash", str(HOOK)], input=json.dumps(payload),
capture_output=True, text=True, env=env,
)
assert out.returncode == 0, out.stderr
finally:
server.shutdown()
server.server_close()
return seen
def test_hook_sends_every_definition_in_a_write(tmp_path):
repo = tmp_path / "repo"
repo.mkdir()
env = _hook_runtime_env()
subprocess.run(["git", "init", "-q"], cwd=repo, check=True, env=env)
seen = _run_hook_against_sink(tmp_path, {
"session_id": "s-feed-w", "cwd": str(repo), "tool_name": "Write",
"tool_input": {
"file_path": str(repo / "new.ts"),
"content": "export async function onDelete(): Promise<void> {\n"
" const ok = await confirmed({ title: 'x' });\n}\n"
".btn-primary {\n color: red;\n}\n",
},
})
assert seen["path"] == ["new.ts"]
assert seen["shapes"] == ["css:btn-primary,sym:onDelete"]
def test_hook_sends_the_enclosing_definition_for_a_body_edit(tmp_path):
"""An Edit to the inside of onTrash names no definition itself; the hook
must walk the file upward from the edited line and send onTrash."""
import shutil
if shutil.which("tac") is None:
pytest.skip("the enclosing-definition walk needs tac")
repo = tmp_path / "repo"
repo.mkdir()
env = _hook_runtime_env()
subprocess.run(["git", "init", "-q"], cwd=repo, check=True, env=env)
target = repo / "comp.vue"
target.write_text(
"<script setup lang=\"ts\">\n"
"async function onTrash(): Promise<void> {\n"
" const ok = await confirmed({ title: 'Move to the trash?' });\n"
" if (!ok) return;\n"
"}\n"
"const other = () => {\n return 1;\n};\n"
"</script>\n"
)
seen = _run_hook_against_sink(tmp_path, {
"session_id": "s-feed-e", "cwd": str(repo), "tool_name": "Edit",
"tool_input": {
"file_path": str(target),
"old_string": " if (!ok) return;",
"new_string": " if (!ok) return;\n await guarded(() => store.trashNode(id));",
},
})
assert seen["shapes"] == ["sym:onTrash"]