Postgres reads x.f as the call f(x) when x has no column f. The guard counted
that spelling only after ")." or "].", so over a function in FROM that
returns a single value, or a parameter, the allowlist was bypassed:
"select g.obj_description from generate_series(...) g" returned the comments
on every object of the database, and g.current_schemas, g.txid_status and
$1.ts_token_type passed as well. A refused catalog function after a dot is
now refused too, except the names in COLUMN_NAMES (money, name, oid, version,
xid, xml), which take no argument or only convert it to their type, so t.name
keeps working. PGPLUG-021 holds that list to pg_proc on the lane's server.
The type aclitem names roles: '=r'::aclitem printed the cluster's first
superuser and 'x=r/x'::aclitem told whether a role exists. It is refused in
every spelling like the reg* types.
The allowlist itself moves to assets/postgres/plugin-functions.txt, beside
the catalog list it is checked against. As TypeScript it spelled 'like' and
'json_each', names of Postgres functions that the SQLite dialect gate reads
as SQLite SQL.
Since the lane runs every suite that boots the app it takes about ten
minutes on Postgres 18 and about 25 on 14, past the job's 20 minute timeout.
Each major now runs it in three slices (--shard=i/3), each on its own runner
and service with its own template; the probe and its artifact stay once per
major, in the first slice. ci-ok still needs server-postgres, whose result is
failure when any slice fails or is cancelled, so every slice is required.
The wiki's move section, the server guide and the nul-free header said
the move drops NUL as every write does. They now say what it does: a
plugin's data moves unchanged or the move is refused with the values
listed, TREK's own NUL is refused unless --drop-nul is given, which lists
every value it changed, and the checksum comes from the SQLite values.
nulFreeInput left a string as sent under any key containing "password".
That covered the admin settings password_login and password_registration,
which are flags: 'true\u0000' failed the lockout check as a disabled
password login while the database stored 'true'. It also tested the raw
key, so 'pass\u0000word' lost the NUL of the password it carried.
The exemption is now HASHED_PASSWORD_KEYS, the fields TREK only ever hashes
with bcrypt (password, current_password, new_password), matched exactly on
the key without its NUL. A unit test reads every field of the shared
schemas and the instance settings named like a password, passphrase or
secret and fails until each is either in that list or decided against with
its reason.
The plugin SQL guard refused functions by name, so every function that reads
the catalogs or the server and was not listed passed: obj_description,
col_description and shobj_description returned the comments on TREK's
tables, columns, database and roles by OID, format_type named a type by its
OID, and version(), current_database(), inet_server_addr() and the session
keywords described the server.
A call to a pg_catalog function is now refused unless the function is on
PLUGIN_FUNCTIONS (string, number, date and time, JSON, array, aggregate,
window, UUID, range, geometry, network and text search functions), in
either spelling of a call, f(...) and the column notation (x).f. The
catalog's function names of Postgres 14 to 18 are kept in
assets/postgres/catalog-functions.txt, and PGPLUG-021 fails on the lane
when the server has a function that list does not name, so a new major
cannot widen what a plugin may call. The session keywords current_user,
session_user, current_role, user, system_user, current_catalog and
current_schema are refused unquoted.
The plugin-sdk README, the Postgres wiki page and server/CLAUDE.md now name
the layers as the code enforces them: the refusal on a privileged role and
the boot log line, the guard (the reg* types, INTO after the statement's
own INSERT or MERGE only, TEMP only where it makes a temporary object), the
targets Postgres resolves before an ALTER, DROP, REFRESH, TRUNCATE or
CREATE INDEX, one statement per round trip, pg_temp last, and the quota over
every relation with the object cap. They also say that a plugin argument
holding NUL is refused on Postgres and bound exactly on SQLite.
The privileged-role check missed three ways a role can reach the database
host from SQL. Before PostgreSQL 16 a member of a CREATEROLE role sets its
role to it and grants itself pg_execute_server_program, since role
attributes are not inherited but the grant checks the role in effect. A role
granted EXECUTE on lo_export, lo_import, pg_read_file, pg_ls_dir and their
siblings reads or writes server files, and lo_export keeps the file after the
transaction rolls back. From PostgreSQL 15 on a role can be granted ALTER
SYSTEM on a setting or SET on a superuser-only one.
PRIVILEGED_ROLE_SQL now counts all three. The parameter grants are read
through query_to_xml inside a CASE taken only on 15 or later, because a
statement naming pg_parameter_acl does not parse on 14. A SET grant on a
setting every role may change anyway, and pg_monitor, stay ordinary. The
lane test creates each kind of role and shows the escalation the check is
for on the server it runs against.
The database stores no NUL on either engine, but the application still held
the value a client sent. A key or a check made from it was then a different
value from the one the database compared: `victim@x.test` followed by NUL
was a fresh key for the per-email password-reset throttle while the lookup
found the victim's row, so every spelling mailed a new link and voided the
last one.
Every input now loses its NUL where it enters TREK: the nulFreeRequest
middleware after the body parsers (path, query string, body, cookies), the
first global pipe for what a route parses itself (a multipart body), the
websocket adapter for every frame, and the MCP registry through a new
host-supplied input mapper, since /mcp keeps its raw body. Headers need
nothing: Node refuses a NUL in one with a 400. A string under a key naming a
password stays as sent: it is only ever hashed, and stripping it would lock
out an account whose password was set with one. The stripping in the
database layer stays as the backstop for what does not come from a request.
The guard's header, server/CLAUDE.md and the Postgres wiki page named
extended-query.ts and a hook on every pool connection. They now say that the
store sends each plugin statement over the extended protocol on a client it
checks out per call, and drops that client's temporary tables and session
state before handing it back.
The ESLint walls against raw(), the sql tag and SQL text handed to
execute/run/all/get applied to src/db/repositories and src/db/seeders only. A
service, job, guard or MCP tool could send its own SQLite-only SQL through
em.getConnection().execute(...), which the Postgres probe never runs, since it
only calls the repositories. The walls now cover all of src/. The dialect layer,
the engines and the migrations stay the homes of raw SQL, and the four boot-time
lifecycle files that hold SQL text today are named and pinned to their count.
gen:pg-baseline rewrote the baseline with its cut at the newest migration on
disk, so after a migration landed after the cut, running it moved the cut over
that migration while Postgres installs created from the old baseline would
still run it through the migrator. It now refuses while a migration follows the
committed cut, names those migrations and the routine, and moves the cut only
with --move-cut (lint:dialect keeps such a migration under the post-cut rules
either way).
check:pg-baseline stopped comparing SQLite triggers with POSTGRES_TRIGGERS once
the baseline was frozen, so a trigger a later migration created on SQLite went
unnoticed. The comparison now runs in both states, with advice that points at
the migration when the baseline is frozen.
lint:dialect missed several classes of SQL that SQLite reads and Postgres does
not, or reads differently:
- group_concat, ifnull, instr, iif, printf, typeof, total(), changes() and
json_object/json_array/json_valid/json_type/json_quote, as calls in SQL text
and as the bare name handed to eb.fn.
- LIKE in SQL text, Kysely's 'like' operator and MikroORM's $like, which ignore
case on SQLite only. A LIKE that is exact on purpose now says so in a
like-exact comment; the six existing ones (setting keys, storage paths, the
lowercased reference search, sqlite_master) carry it.
- The sqlite member of engineDdl was not read at all, so a trigger or a data
backfill put there never reached Postgres. It may now hold only the schema
change and a rebuild's copy and sqlite_sequence bookkeeping, and SQLite
statements with an empty postgres member need a sqlite-only reason.
- execBestEffort, bestEffort and execUnlessTableMissing in a post-cut
migration: on Postgres a swallowed failure aborts the migration transaction.
- The post-cut rules follow POSTGRES_FIRST_CUT as well as the manifest, so
regenerating the baseline over a new migration no longer releases it.
src/db/engine/sqlite/ is SQLite's own code and is no longer scanned.
The gate matched only isSqlite(/isPostgres(, instanceof on a platform class
and a comparison on a property named engine or id. A destructured engine name,
the literal on the left, a switch over the engine, instanceof on a driver class
and a comparison with a driver class all passed with an empty baseline, and
standalone-orm.ts already held one such check. The regex now matches every one
of them, the guard in standalone-orm.ts moves into the engine port as
assertSqliteOrmConfig, and the advice names where a check belongs.
server/CLAUDE.md names where NUL is dropped, INTCOL and the .int() rule
for integer columns, and the calendar-date check for a bound date, and no
longer lists the places NUL suite as SQLite-only. The Postgres wiki says
the move drops a NUL instead of refusing it.
The Postgres wiki page and the plugin author guide list what the plugin
SQL guard now refuses and the two layers behind it, and the wiki and
server/CLAUDE.md name every kind of role a restore refuses, including
CREATEROLE on PostgreSQL 14 and 15, which the managed postgres role on
such a server holds.
They boot on a SQLite file the retired positional runner built, which no
Postgres database ever came from, so they belong in SQLITE_ONLY rather
than in the snapshot baseline. The lane scan now tells such a suite apart
from the snapshot mock: it mocks src/db/database onto a database of its
own, never names createSnapshotTestDb, and is named in SQLITE_ONLY with
its reason. A snapshot suite added to SQLITE_ONLY stays in the baseline.
The plugin-sdk README says what dialect a plugin writes when TREK runs on
Postgres, what the guard refuses and how the limits hold there. The wiki's
Postgres page describes the plugin schemas, their backup and restore and
the one-instance-per-database limit, and its key rotation path now names
the file where it is.
- Database-Postgres: supported versions, the version 18 client and the tool
search order, sslrootcert with a client before 16, how to create a role
that is not a superuser and move an older compose install to one, a
current schema other than public, and the space a restore needs.
- Updating: the manual rollback clears the whole schema and loads the copy
in one transaction, as TREK's restore does; dropping only the tables left
the collation, the trigger functions and the sequences behind. A unit
test keeps the printed statement equal to CLEAR_SCHEMA_SQL, and the boot's
rollback hint points at that section.
- Troubleshooting, Install-Docker-Compose, Environment-Variables, README
and server/CLAUDE.md follow.
TREK runs on PostgreSQL when TREK_DB_URL is set, and on SQLite as before
otherwise: an engine port under src/db/engine, the dialect helpers under
src/db/dialect, a Postgres baseline for the migrations up to the cut,
pg_dump/pg_restore backups, a probe of every repository method against a
real server and a Postgres test lane in CI.
The pre-migrate snapshot docstring said an older release refuses a database a newer TREK migrated. Only images from 5.0 on do; 4.3.x and earlier start on it without a warning. server/CLAUDE.md said anything the legacy baseline cannot map refuses the boot, which holds only above step 258.
A response-contract-exempt comment counted any handler as covered, whatever it returned, so one comment line let a JSON handler past the ratchet. The marker now counts only on a handler taking @Res() or @Next(). The five vacay holiday proxies return the provider's JSON, so they lose their markers and are counted again until they declare a schema.
The ratchet only read *.service.ts for status literals and one branch shape in controllers and tools, so an envelope in a helper, impl or provider file, a branch written as result.error === 'not_found', and RPC handlers all went uncounted. It now counts the literals in every file under src/nest outside the HTTP layer, and comparisons, negations, optional chains and ternaries on a returned error in controllers, tools and RPC handlers. The baseline takes what the wider count finds.
src/nest-rpc, src/nest-mcp and src/mcp import domains and are imported by them, but the domain map only knew src/nest, so their edges dropped out of domainCycles and their repository injections out of foreignRepositories. Each layer now counts as one domain. The repository rule reads the syntax tree for @InjectRepository(X) and for getRepository(X) on an EntityManager, so a comment no longer counts and a foreign read through the EntityManager no longer passes. The baseline takes the cycles and the repository reaches the wider view finds; settings -> Users goes, it was a comment.
sync_trip_documents and the five manual school-holiday tools carried no demo check of their own before the gate moved into the registry, so a demo account could call them. The gate now lets exactly those through, and a test pins the list. upload_trip_file keeps the gate's order: a demo account is refused before its trip access is checked, where the tool used to answer a trip it cannot see with the no-access result first.
lint:test-mocks counts vi.mock/vi.doMock of src/config, src/nest/realtime and src/db per file under tests/, lint:test-new-service counts new XService(/new XRepository( there with every class whose constructor takes nothing allowed (derived from src/ and tests/helpers/), and lint:size holds files under tests/ to 2000 lines against their own baseline. Both counters share scripts/lib/count-ratchet.mjs, which gains a prepare hook for a check that reads the tree first. The three run as steps of the existing server lint job; the CLAUDE.md files and the module guide describe the test helpers they point to.
A hand-built service gets a FakeRealtimeService where it took a RealtimeService, a Nest testing module overrides the provider, and a suite over buildApp() spies on the container's instance (spyOnRealtime). The 167 vi.mock factories on src/websocket are gone, RealtimeService reads the socket registry in ws-state.ts directly, and the re-export stub src/websocket.ts is deleted. The facade's own suite now checks what reaches a socket instead of which module function it called.
The decorator records the @trek/shared response schema a route answers
with. Under NODE_ENV=test it adds an interceptor that parses every response
and fails the request with a logged 500 naming the handler and the issues,
so the e2e and integration suites catch a response that drifted from its
contract; in production nothing is added to the request path and the body
is never rewritten. lint:response-contracts holds the handlers still without
one to a per-file baseline that only shrinks, and runs in server-quality.
toMinor, sumMinor, splitEqualShares and currencyDecimals move to
shared/src/money. The budget service and its MCP tools call them instead of
private copies, and the share table both sides were pinned to becomes the
spec of the shared function. Stored values and every computed figure stay
the same.
A domain injecting the repository of a table another domain owns couples
the two below the service layer, and nothing counted it. The new
foreignRepositories rule reads every @InjectRepository under src/nest,
looks the entity up in scripts/repository-owners.json (an injected entity
without an owner stops the run) and holds each domain/entity pair to the
baseline, which only shrinks: 209 pairs today.
Every write tool used to open with its own demo check, 196 copies across 30
tool files, most of them injecting AuthService for nothing else. The
registry now takes a toolGate (built from the container) that runs before
each registered tool handler, and TREK's gate refuses a demo account on
every tool not annotated read-only with the same canned result. The
per-tool checks, the AuthService injections, AuthService.isDemoUser and the
AuthModule imports that only fed them are gone; ESLint refuses a demo check
in a tool file, and a registry-wide test calls every write tool as the
demo account.
app-settings.registry.ts lists every app_settings key the server reads by name with its shape, default, environment override and flags. The admin form's keys, its masked and encrypted handling and the public config's reads derive from it, isOidcConfigured() replaces the three issuer/client-id checks in AuthService, and OidcService resolves its values through the same rule while still requiring the client secret. Every literal getValue() read goes through readAppSetting()/resolveAppSetting(), and ESLint refuses a new one.
server/CLAUDE.md now says which API new repository code uses: Kysely for join, aggregate and set-based statements, the entity methods for single-table CRUD, no new QueryBuilder calls. The TrekRepository docstring no longer calls Kysely an escape hatch. lint:query-api counts this.qb()/createQueryBuilder() per repository file against a baseline that may only shrink and runs in the server CI job.
The same file documents the JSON column and timestamp rules from the two changes before.
tsconfig.strict.json is the build config with strictNullChecks and
noImplicitAny on. lint:strict runs tsc over it and holds the errors per file
to scripts/strict-baseline.json (204 in 109 files today): a file may only
lose errors, a new file starts at none, and an entry above its count fails
until --update lowers it. An error outside any file or an unreadable tsc run
fails closed. It runs in the server job; the build itself stays on
strict: false.
shared's lint ran eslint --fix with prettier/prettier as an error, so in
lint-prettier.yml the format:check after it only ever saw files it had just
rewritten, and the server's format scripts pointed at a test/ directory that
does not exist. Now lint is check-only in both (lint:fix applies fixes), the
server scripts cover src/ and tests/, and a lint:format ratchet on the shared
library holds every file outside scripts/format-baseline.json to Prettier in
both workspaces: in the server job of test.yml and in lint-prettier.yml,
which keeps providing the lint check. Formatting leaves ESLint in shared, and
the unused eslint-plugin-prettier goes. Once the tree is reformatted in one
commit, --update empties the baselines and the check is a full prettier
--check.
Server lint ran without --max-warnings or a baseline, so the rules left on
warn could grow without limit. lint:warnings runs ESLint with the server's
own config through the shared tally, fails on any error or inline rule
config, and holds the warnings per rule for src, tests and eslint-disable'd
messages to scripts/eslint-baseline.json, which only falls. It replaces
lint:check in CI. no-await-in-loop, no-non-null-assertion and no-console come
in as warnings held at today's counts. npm run lint no longer rewrites files;
lint:fix does.
acquire() made sure a row existed and then raced on a conditional UPDATE,
two statements the registrar wrapped in a transaction of its own. lint:tx
cannot follow a repository reached through a local variable, so it
reported the method. One INSERT ... ON CONFLICT DO UPDATE ... WHERE now
takes a free lease, renews the holder's and leaves anyone else's alone,
atomically on SQLite and on Postgres, and needs no transaction.
The server suite seeds and reads its database through some 5,500 raw
better-sqlite3 statements on the sync handle. They restate the schema by
hand, skip the repositories the app goes through and run on SQLite only,
so nothing in the suite could ever exercise the dialect layer on another
engine.
tests/helpers/factories/ adds typed factories per domain (users, trips and
days, places, assignments, bookings, budget, packing, to-dos, files,
journeys, tours, collections, notifications, settings, plugins, tokens,
atlas, vacay, photos, collab) and generic readers (findRow, findRows,
countRows, updateRows, deleteRows) that go through MikroORM in a request
context of their own, against the app's ORM or a createTestOrm handle.
They refuse a write to a read-only foreign-key twin, which MikroORM would
otherwise drop without a word. The README there is the conversion guide.
The categories integration suite and the tags and calendar-feed e2e suites
are converted in full as worked examples; the tags suite also moves from
hand-rolled tables to the migrated snapshot.
lint:test-sql (scripts/test-sql-ratchet.mjs) holds the raw statements left
to a per-file baseline: growth and stale entries fail, files whose subject
is raw SQL are exempt with a reason, and a single statement can be allowed
with a test-sql-allow comment. CI runs it next to lint:boundaries.
A session renewed from a token issued before sessions were tracked has
an id derived from that token, and the renewal inserts it only when no
row by that id exists. The nightly purge deleted revoked rows straight
away, so the next request with the still valid old token inserted the
session again as an active one and got a fresh tracked token with a new
lifetime. A logout, an ended session or a sign out of the others held
only until the purge ran.
The purge now deletes expired rows only, and a revoked row stays until
its own expiry. A derived row is recorded with an expiry no earlier
than the old token's, so by the time it goes the old token is refused
by its signature check anyway.
The token in the sign-in bodies was declared deprecated and "read by
nothing of ours", but the web client still decided an SSO sign-in by
reading data.token from GET /api/auth/oidc/exchange, an endpoint the
deprecation notice left out. Removing the field later, as documented,
would have broken SSO sign-in.
The exchange now answers { success: true, token } and the login hook
signs in when the response is OK and success is true, so it never reads
token. The field stays for API clients and is now on the deprecation
list in the API docs, the controller comments, server/CLAUDE.md and the
wiki (Security-Hardening, OIDC-SSO). Adding a field to the body breaks
no existing client.
A session token was a signed JWT and nothing else. Logout only cleared
the cookie, a 30-day "remember me" token stayed valid wherever a copy of
it had ended up, and nobody could sign out one device short of changing
the password or rotating the JWT secret for everyone.
Every session token now carries an id (jti) naming a row in the new
user_sessions table (id, user_id, created_at, last_seen_at, expires_at,
revoked_at, user_agent cut to 256 characters; no address is kept). The
rows live behind UserSessionsRepository and the new sessions domain,
whose SessionsService is the one issuer: AuthService.generateToken and
the OIDC sign-in both go through it, and sliding renewal re-signs a
tracked token under the same id and moves its expiry. verifyJwtAndLoadUser
takes the repository and refuses a token whose session is revoked,
expired or gone, refreshing last_seen_at at most every five minutes.
A token without a jti was issued before this change; it passes until its
own expiry (or a password change), so the upgrade signs nobody out, and
renewal turns it into a tracked session.
Sessions end on logout (the one the request names, verified even when
expired), on a password change (all, the device gets a new one), a
password reset, an admin setting a password, reset-admin.js, an MFA
disable (all but the current one), an admin clearing a user's MFA, and
with the account. GET /api/auth/sessions lists the active sessions,
DELETE /api/auth/sessions/:id ends one and POST
/api/auth/sessions/revoke-others ends all but the current, with their
contracts in @trek/shared and two new audit actions. There is no UI for
them yet. MCP exposes no account or session tools, so there is no
parallel tool to change; an MCP call made with a session JWT as bearer
goes through the same check.
The token in the login, register, demo-login, MFA and passkey bodies
stays for API clients but is marked deprecated in the API docs and in
the client type, which never read it.
A nightly session-purge job on CronRegistrarService deletes the expired
and revoked rows. The migration is 042500. The sessions domain gets its
own coverage entry at the 80 floor until a full run measures it.
The rule that a multi-statement write runs in uow.transactional was
written down but nothing held it. scripts/tx-writes.mjs reads every class
under src/ with the TypeScript parser and counts, per method, the writes
it issues outside a transactional callback: repository calls that build
an INSERT, UPDATE or DELETE, calls to methods that write (on the class
itself or on a constructor-injected one), and a transaction holding
writes as one. A write in a loop counts twice, of two branches only the
larger counts, local helpers count where they are called, and a helper
whose every call sits inside a transaction is not reported. --explain
lists the line of every call that counted.
Today's 102 reported methods sit in scripts/tx-writes-baseline.json and
may only shrink; most are syncs with network I/O between their writes,
plugin lifecycle steps and adapters calling two service writes. A new
method fails CI, and an entry above what its method holds fails until
--update lowers it, the same ratchet as lint:size and lint:boundaries.
Two JSDoc tags mark the deliberate cases with their reason in the tag:
@txStandalone for a write that stands on its own whatever the caller does
(the audit row, notification sends, the booking price mirror, the
assignment journey refresh, the quota retention purge), @txIndependent
for writes complete one by one on purpose (the OAuth token retention
sweep, the notification claim and its release). Runs in CI after
lint:boundaries; documented in server/CLAUDE.md, the root CLAUDE.md and
the wiki's dev setup page.
The time budget covered query() and tx() only. exec() without bound args
handed the plugin's script to Database#exec, which runs any number of
statements in one call with nothing checked between them, so a long
script held the host loop as long as it liked.
An exec script now runs one statement at a time. sql-script.ts cuts it
after every semicolon outside a string, a quoted identifier or a
comment; a piece SQLite reports as incomplete input (a trigger body cut
at its inner semicolon) is joined with the next until it prepares. The
clock is read after every statement, and past the budget the call
throws. As with Database#exec, the statements before that point stay
applied, and an error is SQLite's own; a transaction the script opened
itself is rolled back so the connection is not left inside it.
exec() with bound args is one statement and stays as it was. A single
statement cannot be bounded at all on the host thread, a write such as
INSERT ... SELECT over a cross join included, and the SDK README and the
wiki now say so next to the cross-join read. migrate() keeps running
unbudgeted on purpose: a migration stopped part of the way is rolled back
and fails the same way on every later start, leaving the plugin unable to
update its schema.
The ctx.db rows of the SDK README and the plugin wiki pages describe the
budget for scripts, and the lines touched there and in plugin-sdk's db
comment lose their em dashes.
The output contract cut only the top level of an entity result. A derived
key's value went out untouched, so the children that are whole rows still
carried every column: a day's notes_items are full day_notes entities and
a reservation's endpoints are SELECT e.* on reservation_endpoints. About
fifty methods were also left as readModel and passed through whole, many
of them backed by SELECT * reads (day_accommodations, day_notes, todos,
tags, packing bags, journal entries, collab, collections, vacay plans).
A new column on any of those tables reached every plugin by default,
which is what the contract was meant to stop and what the SDK README and
the wiki already claimed it did.
PLUGIN_ENTITY_NESTED now names, per entity, the derived keys that hold
rows of another entity, and the shaper cuts those rows down to their own
field list, recursively. dayNote and reservationEndpoint are the child
tables that were whole rows; day.assignments and place.tags are held to
the assignment and tag lists too, although their reads already name
their columns. The remaining derived children (travelers, participants,
ratings, budget members, payers and receipts, packing recipients,
attachments, voters, bag members) are built field by field, which the
contract comment now states as the rule for a derived key without a
nested entity.
The whole-row read models get entity contracts: accommodation, dayNote,
todo, tag, category, packingBag, fileLink, collabNote, collabPoll,
collabMessage, journey, journalEntry, bucketItem and collection, and
vacayPlan with its holiday calendars. collections.listMine,
collections.get and vacay.mine return envelopes around rows; an entity
with a null table describes those, so their rows are reached through the
same nesting. Every list holds exactly the keys each read returned
before, so no plugin loses a field. What stays readModel is built field
by field from named columns, and each such entry says what it returns:
setBagMembers, atlas.visited, collections.savePlace and copyToTrip, the
two vacay toggles and journal.addEntryPhoto.
The schema test now covers every child and new table, and calls each
newly shaped method on the real services, comparing nested rows key for
key so the only field allowed to disappear is a withheld one. A further
case proves the nested rows it compares are not empty. gen:plugin-facts
publishes PLUGIN_ENTITY_NESTED to the SDK, and the SDK README, the plugin
wiki pages and server/CLAUDE.md describe the nested rows and list the new
entities.
A plugin's own database runs on the host's event loop, and the only bounds
were a regex denylist, the 256 MB quota and a 100,000 row cap. A select
whose every row is expensive, or a long scan under the cap, could hold the
loop for as long as it liked.
query() and tx() now get a wall-clock budget of 2 s, read between the rows
a statement yields and between the statements of a batch. Past it the call
throws, and a tx rolls back whole like any other failure inside it. The
row cap and the budget are overridable per PluginDataDb for tests; the
defaults are what production runs.
better-sqlite3 has no interrupt or progress handler, so a statement that
yields nothing until it is done, such as an aggregate over a cross join,
still runs to its first row. Only moving plugin SQL off the host thread can
bound that; the SDK README and the plugin wiki state the budget and this
limit.