Repository navigation
Conversation
This was referenced Oct 16, 2026
ditadi
force-pushed
the
stack/database-dev-flow/05-db-init
branch
from
October 19, 2026 20:35
960b211 to
c56de21
Compare
`appkit db init` gives a checkout its own Lakebase branch, forked from the project's default branch: your dev branch, or with --session a branch for this checkout that expires after 7 days (or --ttl). It asks nothing. It refuses the default branch, protected branches and branches databricks.yml deploys to before any change, and when the project is at its branch limit it lists your branches to delete. It makes sure the branch has a read-write endpoint that suspends after 5 idle minutes and a Postgres role for you, points .env at it with an atomic 0600 write, hands you the app schema when the fork copied it owned by another Lakebase role, installs drizzle-kit when the app lacks it, then starts the schema by migrating, introspecting or cleaning. `appkit db clean` drops the app schema (only your objects in public) and lists what outside it went along; --delete-branch deletes this checkout's session branch instead. `appkit db seed` runs config/database/seed.sql in one transaction with the app schema on the search_path. Both run only on a branch created for you or this checkout. Signed-off-by: ditadi <victordperd@gmail.com>
ditadi
force-pushed
the
stack/database-dev-flow/05-db-init
branch
from
October 20, 2026 12:22
c56de21 to
98ef27b
Compare
Contributor
🤖 AppKit PR bot🔬 Run evalsStart an eval for this PR from the evals-monitor app: Go to Evals Monitor → 📦 Try this PR's app templateScaffolds a new app from this PR's SDK build. Run it in any folder (requires the GitHub CLI — gh run download 41109404809 -R databricks/appkit -n appkit-template-0.89.0-pr.aa5285d-stack-database-dev-flow-05-db-init-672 -D appkit-pr-672 \
&& unzip -o "appkit-pr-672/appkit-template-0.89.0-pr.aa5285d-stack-database-dev-flow-05-db-init-672.zip" -d "appkit-pr-672" \
&& databricks apps init --template "appkit-pr-672"The template pins |
This was referenced Oct 20, 2026
Contributor
📦 Bundle size reportCompared against
|
| dist | raw | gzip |
|---|---|---|
| JS (runtime) | 1.4 MB (+78 B) | 515 KB (+4 B) |
| Type declarations | 478 KB | 175 KB (+2 B) |
| Source maps | 2.8 MB (+317 B) | 958 KB (+88 B) |
| Other | 11 KB | 3.7 KB |
| Total | 4.7 MB (+395 B) | 1.6 MB (+94 B) |
Per-entry composition (own code — deps external (as shipped))
| Entry | Initial (gz) | Lazy (gz) | Total (gz) | node_modules (min) | Own code (min) |
|---|---|---|---|---|---|
. |
106 KB | 2.5 KB | 108 KB | external | 343 KB |
./beta |
108 KB | 485 B | 108 KB | external | 323 KB |
./beta/database-tooling |
45 KB (+16 B) | 0 B | 45 KB (+16 B) | external | 134 KB (+50 B) |
./testing |
45 KB | 32 KB | 77 KB | external | 223 KB |
./tsdown |
520 B | 0 B | 520 B | external | 813 B |
./type-generator |
26 KB | 0 B | 26 KB | external | 74 KB |
Chunks:
| Entry | Chunk | Load | Size (gz) |
|---|---|---|---|
. |
index.js |
initial | 101 KB |
. |
utils.js |
initial | 4.6 KB |
. |
remote-tunnel-manager.js |
lazy | 2.5 KB |
./beta |
beta.js |
initial | 88 KB |
./beta |
modular.js |
initial | 6.0 KB |
./beta |
stream-manager.js |
initial | 5.9 KB |
./beta |
service-context.js |
initial | 4.2 KB |
./beta |
databricks.js |
initial | 3.3 KB |
./beta |
client.js |
initial | 647 B |
./beta |
index.js |
initial | 20 B |
./beta |
supervisor-api.js |
lazy | 192 B |
./beta |
databricks.js |
lazy | 178 B |
./beta |
index.js |
lazy | 115 B |
./beta/database-tooling |
index.js |
initial | 45 KB |
./testing |
manifest.js |
initial | 32 KB |
./testing |
index.js |
initial | 10 KB |
./testing |
wide-event-emitter.js |
initial | 2.9 KB |
./testing |
index.js |
lazy | 28 KB |
./testing |
remote-tunnel-manager.js |
lazy | 2.5 KB |
./testing |
utils.js |
lazy | 1.8 KB |
./tsdown |
index.js |
initial | 520 B |
./type-generator |
index.js |
initial | 26 KB |
@databricks/appkit-ui
npm tarball (packed): 397 KB (+4 B) — gzipped download (dist + bin; excludes release-only docs/NOTICE).
| dist | raw | gzip |
|---|---|---|
| JS (runtime) | 429 KB | 147 KB |
| Type declarations | 267 KB | 98 KB (+1 B) |
| Source maps | 853 KB | 286 KB |
| CSS | 16 KB | 3.2 KB |
| Total | 1.5 MB | 534 KB (+1 B) |
Per-entry composition (consumer bundle — deps bundled, peerDeps external)
| Entry | Initial (gz) | Lazy (gz) | Total (gz) | node_modules (min) | Own code (min) |
|---|---|---|---|---|---|
./js |
5.3 KB | 49 KB | 55 KB | 208 KB | 14 KB |
./js/beta |
2.1 KB | 0 B | 2.1 KB | 0 B | 5.0 KB |
./react |
433 KB | 49 KB | 481 KB | 1.3 MB | 177 KB |
./react/beta |
4.9 KB | 0 B | 4.9 KB | 0 B | 12 KB |
Chunks:
| Entry | Chunk | Load | Size (gz) |
|---|---|---|---|
./js |
index.js |
initial | 5.2 KB |
./js |
chunk |
initial | 120 B |
./js |
apache-arrow |
lazy | 49 KB |
./js/beta |
beta.js |
initial | 2.1 KB |
./react |
index.js |
initial | 431 KB |
./react |
tslib |
initial | 2.1 KB |
./react |
apache-arrow |
lazy | 49 KB |
./react/beta |
beta.js |
initial | 4.9 KB |
Signed-off-by: ditadi <victordperd@gmail.com> # Conflicts: # packages/appkit/src/database/tooling/migrate.ts # packages/appkit/src/database/tooling/naming.ts # packages/appkit/src/database/tooling/tests/guard.test.ts # packages/appkit/src/database/tooling/tests/naming.test.ts # packages/appkit/src/database/tooling/tests/targets.test.ts
init still checked its planned branch against the readDeployBranches list that the deploy boundary replaced. It now asks deployRefusal, like every other guard, so a binding the checkout's files cannot resolve refuses init before any API call or .env write. readDeployTarget reads the default target from the same bundle model: its first resolved app binding, else its postgres_branch and postgres_database variables. The shared CLI test mocks the beta tooling path. Signed-off-by: ditadi <victordperd@gmail.com>
Signed-off-by: ditadi <victordperd@gmail.com>
clean drops the branch's migration journal, and init runs clean, migrate or introspect after its own checks, so another appkit db command could run in between. runClean now holds the checkout's history lock, and init holds one lock for its whole run and hands it to cleanHolding, migrateHolding and introspectHolding. init --schema none changes no history, so it takes no lock and creates no config/database. Signed-off-by: ditadi <victordperd@gmail.com>
init --schema none still handed the app schema to the developer, which deletes the branch copy of the role that owns it, in apps that may have no database schema contract at all. It now leaves the schema and its owner as the fork copied them. Schema public is never handed off either: it belongs to the database (pg_database_owner on PostgreSQL 15 and later, the bootstrap superuser before), so AppKit never deletes its owner or reassigns its objects. Signed-off-by: ditadi <victordperd@gmail.com>
Regenerated with pnpm clean && pnpm build && pnpm size:baseline. Against the previous baseline the @databricks/appkit packed tarball grows from 1364290 to 1448119 bytes (+83829, +6.14%), over the 5% budget. The growth is the CLI-only @databricks/appkit/beta/database-tooling entry (45 KB gzipped, from 28 KB) with the introspect fixes and init, clean and history lock code it ships. @databricks/appkit-ui is unchanged at 406741 bytes. Signed-off-by: ditadi <victordperd@gmail.com>
The re-run test ran with --schema none, which skips the hand-off, so no test reached the check that keeps init from deleting your own role. It now runs the default migrate start and asserts that the owner query ran for schema app and no role was deleted. Signed-off-by: ditadi <victordperd@gmail.com>
init --schema none hands off no schema and starts none, but it still loaded schema.ts for the schema name, so a half-edited file stopped a branch-only init. It now loads schema.ts only when the hand-off and the start use it. Signed-off-by: ditadi <victordperd@gmail.com>
…e lock withHistoryLock creates config/database for its lock file, and init and clean checked for schema.ts only after taking it, so a refused init or clean left an empty config/database behind. Both now check first, as runMigrate does. Signed-off-by: ditadi <victordperd@gmail.com>
dotenv 16.6.1 reads `KEY: value` (a colon followed by whitespace) as an entry, and a later entry wins. updateEnvFile's parser only knows `KEY=`, so a later `PGHOST: old` outlives the rewrite and dotenv reads the old host. Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
updateEnvFile only recognized `KEY=`. dotenv 16.6.1 also reads `KEY: value` when whitespace follows the colon, and its last entry wins, so a later `PGHOST: old` outlived the rewrite. The parser now treats that form as an entry too and writes it back as `KEY=`. Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
…nv through updateEnvFile writes `..env.<pid>.<hex>.tmp` and renames it over .env. A run killed between the two leaves the file, which holds the .env contents, and nothing ignored it. Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
init kept any existing PGSSLMODE. The Lakebase pool maps prefer to rejectUnauthorized: false, disable to no TLS, and rejects every other value, so only require gives a checked TLS connection. init now writes require over anything else and says what it replaced. Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
Co-authored-by: Isaac <no-reply@databricks.com>
The fork's parent is the branch the default target binds when the project has it, else the project's default branch, and the step names which. Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
…s branch init forked the project's default branch, but the branch databricks.yml deploys to is the one that holds the deployed migrations. When the default target binds a live branch of the same project, init forks that one and falls back to the project's default branch otherwise. The step and the report name the parent and which rule picked it. Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
…t none Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
init installed drizzle-kit only for the migrate and introspect starts. A new fork's default start is the rebuild, so it skipped the install, and the next `appkit db migrate` or `appkit db check` failed to load drizzle-kit. Every start except none leaves the branch for those commands, so every start except none installs it. Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
… .env marker APPKIT_DB_NEW_BRANCH recorded in one checkout's .env that init forked a branch and had not finished its first start. It was checkout-local while the dev branch is per user, a stale one could drop a developer's work without a prompt, and --schema none, migrate or introspect cleared it while keeping every copied row. It is gone. init now reads the branch: the app schema is absent, the fork's untouched copy (the service principal owns it), a copy handed to you whose rebuild stopped (you own it and its journal carries the appkit:fork-copy row), or your working branch. One function maps that state and the --schema request to the start. A copy is tagged, handed off and rebuilt without a prompt. A tagged copy is rebuilt on any rerun, from any checkout, without another hand-off. Your working branch is migrated, and an explicit clean of it asks first through a confirmation the caller supplies. --schema migrate or introspect on a copy hands it off, removes the tag, keeps the rows and reports where they came from. --schema none touches nothing. public is never handed off or tagged: a new fork is migrated and the service principal's tables that kept their rows are reported. Without schema.ts, introspect writes the files before the tag and the hand-off, so a refusal leaves the copy as the fork made it, and records nothing on a branch the rebuild drops right after. appkit db migrate, introspect and seed refuse a branch whose journal carries the tag. Covered end to end against Postgres 17 with the Lakebase roles modelled. Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
…, and report kept copied rows The confirmation for --schema clean moves from before the run into a callback init calls once it has read the branch, so a fork's copy is rebuilt without a question and a non-interactive shell is refused only when there is one to ask. A no prints what init left and exits 1. init's report of rows the fork copied that stayed, on request or because other roles own them in public, is printed as a warning. The option and help text describe the rule the branch decides. Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
…ng branch Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
…e schema Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
# Conflicts: # packages/appkit/src/database/migrations/ownership.ts # packages/appkit/src/database/migrations/tests/ownership.test.ts # packages/appkit/src/database/tooling/migrate.ts
…'s copy Four Postgres regressions from the review of design 001, failing before their fixes. The fork-copy tag must not be written into a journal owned by a role the migrator does not trust. A fork this run created is a copy even when you already owned the source schema, so init rebuilds it. migrate, introspect and seed refuse an untouched copy as well as a tagged one. After keeping the copied rows, a plain init keeps them and only --schema clean removes them, which the CLI warning now says. Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
…rk copy They checked only the fork-copy tag, so a copy the service principal still owned, with no tag yet, took writes. appkit db seed could add rows on top of the copy through DATABRICKS_SUPERUSER, and the next init dropped them. The three commands now read the schema's state and refuse a copy as well as a handed-off copy. The refusal names the owner for an untouched copy and the stopped rebuild for a tagged one, and gives the same advice for both. init's rebuild still reads a copy with record: false, since that writes nothing on the branch. Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
tagForkCopy, untagForkCopy and isForkCopy checked only that an ordinary table stood at the journal's name. The migrator refuses to record into a journal owned by a role it does not trust, and to read one under row-level security, because a trigger, rule, default or policy there would run that role's code as the caller. The tag helpers now reuse that rule through readJournalRelation and journalRefusal, exported from migrations/apply with its behavior unchanged. A fork's untouched copy keeps the schema owner's journal, which the rule trusts, so the tag still works there. Where it does not, init rebuilds without a tag. Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
…d it on the source readSchemaState returned yours for an untagged schema you own, and planStart migrated it, so a new fork of a source whose app schema you already owned kept the copied rows silently. init knows it forked the branch in this run and now passes that to planStart. A schema on a fresh fork is the fork's copy whoever owns it: the default start tags it, rebuilds it from the committed migrations, and makes no hand-off, since the schema is yours already. Without schema.ts it introspects first, as for any copy. --schema migrate and --schema introspect keep the rows, as they do on a copy. A stop after the tag leaves a handed-off copy the next init finishes; a stop before it leaves a schema that looks like your working branch, the same window as a copy without a journal. Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
…lds it The tag is what lets the next init finish a rebuild that stopped. When the copy has no journal table, or one owned by a role the migrator does not trust, init could not write it and said nothing, while the CLI help promised that the next init always finishes the rebuild. init now warns before the hand-off, through a new onWarning option that the CLI prints as a warning, and tells you to run appkit db init --schema clean if the run stops. The help text says when that promise holds. Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
…removes it After --schema migrate or --schema introspect on a copy, the warning said a plain appkit db init would rebuild the schema without the copied rows. It does not. The tag is gone and the schema is yours, so a plain init migrates it and keeps the rows. The warning now says so and names appkit db init --schema clean, which drops the app schema with everything in it after a confirmation, or with --yes. Signed-off-by: ditadi <victordperd@gmail.com> Co-authored-by: Isaac <no-reply@databricks.com>
The database plugin page and the init help now say that a branch init forked in this run is rebuilt even when you already owned its schema on the source, that the fork-copy row is written only into a journal the migrator trusts and that init warns and tells you to run --schema clean after a stop when it cannot write it, that migrate, introspect and seed refuse an untouched copy as well as a tagged one, and that after keeping the copied rows only --schema clean removes them. Signed-off-by: ditadi <victordperd@gmail.com> Co-authored-by: Isaac <no-reply@databricks.com>
…d the CLI The comment audit of the design 001 fixes flagged narration, type and field docs, and rationale for our own code in this layer's files. This deletes the 64 of them that remain and were written by this fix round, and 11 more the R2-05 unit added since the audit that narrate the same way: the onWarning docs, the record option's caller, test variable docs and comments that restate the assertion below them. The kept comments are the Lakebase why on tagForkCopy and the two Postgres test headers. Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
… of untagging around introspect readMigrationDrift takes ignoreHashes, journal rows that are not migrations, and introspect passes the fork-copy tag. A copy an earlier init tagged before it stopped is then read for init's rebuild as it is, so init no longer deletes the tag before introspect and writes it back after. The behavior stays: introspect on a copy writes files only, and the tag is on the branch before the hand-off. Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
…before the fork The write of APPKIT_DB_SESSION ahead of createBranch is now persistSessionIdBeforeFork, so the order reads from the call instead of a comment: a run that stops during or after the fork leaves its rerun the session id to reuse. Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
…pe instead of a null sentinel
initHolding took LockedStart | null, where null meant --schema none. It
now takes a StartRequest: { kind: "none" } for a run that takes no lock
and starts nothing, or { kind: "locked", lock, requested }. The reads of
schema.ts and of the branch narrow on the kind.
Co-authored-by: Isaac <no-reply@databricks.com>
Signed-off-by: ditadi <victordperd@gmail.com>
The option says what false means: schema.ts and the migration are written, and the migration is not recorded on the branch as applied. init's rebuild passes false, since it drops the schema right after. Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
…e confirm through a predicate The schema state "handed-off-copy" meant a rebuild that stopped after the tag and before its clean, so it is now "interrupted-rebuild". planStart no longer gets confirm: false from a spread default: every start sets it, and an explicit clean asks only where mayHoldYourWork(state, forkedThisRun) holds, which holdsForkCopy defines as neither a copy, an interrupted rebuild, a schema on a branch this run forked, nor an absent one. The created parameter is named forkedThisRun, which is what it says. Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
…med error
A no to the confirmation an explicit --schema clean asks for is a normal
result, not a failure, so runInit now returns start: { kind: "cancelled",
schemaName } with the branch, endpoint, role and .env the run did set
up. InitCancelledError and the CLI's match on its name go. The CLI
prints the same line and exits with code 1 as before.
Co-authored-by: Isaac <no-reply@databricks.com>
Signed-off-by: ditadi <victordperd@gmail.com>
A Rebuild carries reason: "fork-copy" when the branch still held the rows the fork copied and nothing was requested, or "requested" for an explicit --schema clean. The CLI prints its hint about keeping the copied rows from that field instead of inferring it from the start kind and the absence of --schema. Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
…rops its database Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: ditadi <victordperd@gmail.com>
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
🥞 Stacked PR (generated by Stack)
Use this link to review incremental changes.
Stacked on #670. The design is #677.
What did you change, and why?
Each checkout needs its own Lakebase branch for local development, set up without prompts so an agent or a script can run it. This PR adds
appkit db init,appkit db cleanandappkit db seed.appkit db inituses yourdev-<you>-<hash>branch, or with--sessionasession-<user hash>-<id>branch that expires after 7 days (--ttlchanges that). It forks a new branch from the branch the defaultdatabricks.ymltarget binds, or from the project's default branch when that target does not resolve, and says which. It makes sure the branch has a read-write endpoint that suspends after 5 idle minutes and your Postgres role withDATABRICKS_SUPERUSER, and points.envat the branch. It hands you the app schema when the fork copied it owned by another Lakebase role, such as the app's service principal.Then init starts the schema, which it decides from the branch (see the fix round below).
--schema migrate,introspect,cleanandnonechoose a start explicitly. init installs the pinned drizzle-kit for every start exceptnonewhen the app lacks it. init refuses the project's default branch, protected branches and any branch the deploy boundary covers, before it changes anything.cleandrops the app schema, migration journal included, in one transaction. Onpublicit drops only what your role owns.clean --delete-branchdeletes this checkout's session branch.seedrunsconfig/database/seed.sqlin one transaction.cleanandseedrun only on your dev or session branch..envis written with mode 0600 through a temporary file, and a symlinked.envis refused.This revision changes these behaviors:
databricks.ymlreference the checkout's files cannot resolve refuses init before any API call or.envwrite.cleanholds the lock too.clean, and init with--schema migrateor--schema clean, refuse a missingschema.tsbefore taking the lock, so a refused run leaves no emptyconfig/database.init --schema nonereads noschema.ts, takes no lock, and hands off nothing. The schema and its owner stay as the fork copied them.publicis never handed off. It belongs to the database, so AppKit never deletes its owner or reassigns its objects.bundle-size-baseline.jsonis regenerated. The packed tarball grows from 1,364,290 to 1,448,119 bytes (+6.14%), over the 5% budget, because./beta/database-toolinggrows from 28 KB to 45 KB gzipped.Session branch names carry a hash of your user name (#666). init treats
.envas this checkout's session branch only when the name matches your hash, so a.envcopied from another developer never leads init to their branch.appkit dbstays hidden fromappkit --help.Review round, 2026-11-09:
initforks starts without the rows the fork copied:cleanwithschema.ts, orintrospectand then a rebuild from the baseline.--schema migrateor--schema introspectkeeps them; a reused branch is migrated as before. Inpublic, the report lists what other roles own and keep.initrefuses a databasedatabricks.ymlbinds that the fork lacks, reads the deploy target throughvariable-overrides.json, resumes an interrupted session rotation, reconciles with a concurrent run that created the endpoint or role first, writesPGSSLMODE=requireover weaker values, and hands off and introspects the schemaschema.tsdeclares.cleanpreviews the full CASCADE closure; drizzle-kit's installer uses shared's package-manager detection;.envrewrites recogniseKEY: valueand go throughtooling/files.ts; the.envtemp file is git-ignored in the template.Fix round for fjakobs's review of #677, 2026-11-15:
APPKIT_DB_NEW_BRANCHmarker is gone (L236, L690). init reads the branch. An app schema still owned by another role is an untouched copy, and init rebuilds it. Before the hand-off, init tags the copied journal with one row, only in a journal the migrator trusts, so a rerun from any checkout finishes an interrupted rebuild. A branch this run created is rebuilt even when you already owned the source schema. When the copy cannot be tagged, init warns first.--schema nonetouches nothing, so the next plain init still rebuilds a copy. Explicit--schema migrateorintrospectkeep the rows and say that--schema cleanremoves them.migrate,introspectandseedrefuse an untouched or tagged copy.none(L541), and forks the branch the defaultdatabricks.ymltarget binds when it resolves (L633). The hand-off error stops suggestingALTER ... OWNER TO.How do you know it works?
init.test.tsrunsrunInitagainst an in-memory Lakebase. It covers branch creation and reuse, session TTLs, a 403, a 409, the branch limit, and deploy-target and unresolved refusals before any API call. It also covers the hand-off and its failures,--schema none,public, and one lock across init's clean and migrate.env-file.test.ts,clean-seed.test.tsand the CLI'sinit-clean-seed.test.tscover the.envwriter, the clean and seed refusals, the confirmations and the exit codes.clean.pg.test.tsandhand-off.pg.test.tsrun against Postgres. They check what CASCADE takes outside the app schema, apublicclean, and the objects and grants that block a role delete.dbtests passed against a local PostgreSQL 14, typecheck passed, and a clean build matched the new size baseline. The full vitest suite passed at an earlier revision of this PR and at the top of the stack (feat(appkit): add appkit db check #678). The packed tarballs export the same root and./betanames as main, and./beta/database-toolingadds only init, clean, seed and session-branch deletion. CI runs the full suite on this PR.publicon a real Lakebase branch was not observed.Review round, 2026-11-09: each fix has a failing-then-passing test in its PR, and every layer's appkit and shared tests ran against Postgres 14 after the sync. At the top of the stack,
pnpm build,pnpm test(6502 passed, 1 skipped),pnpm -r typecheck,pnpm check,pnpm knipandpnpm docs:buildpass.Fix round for fjakobs's review, 2026-11-15: each fix has a failing-then-passing test in its PR, and every layer's database, plugin and CLI tests ran against Postgres 17 after the sync. A throwaway Lakebase fork confirmed that the developer role can write the service principal's rows but cannot alter its tables. At the top of the stack,
pnpm test(6602 passed, 1 skipped),pnpm -r typecheck,pnpm buildandpnpm docs:buildpass, and oxlint and oxfmt are clean on every changed file.