Repository navigation
v0.7.0-beta.0
Pre-releaseTyped nested data, equality indexes, and in-database TypeScript procedures — plus an everyday CLI for local databases.
This is a beta. Treat it as a new on-disk format, not an in-place upgrade from 0.6.
Breaking: wipe the data directory
0.7 stores USER / SHARED / STREAM row values as compact ordinal payloads (identity stays on the RocksDB key). The server does not dual-decode 0.6 KENV / name-keyed rows.
- Dev and test clusters: wipe the data directory before starting this binary.
- There is no in-process cluster migration in this release. An offline rewrite tool is follow-on work.
- HTTP, WebSocket, PGWire, and SDK payloads are unchanged.
Server functions (#380)
SQL-defined procedures run in a sandboxed V8 isolate, using the same types, row codec, and indexes as tables.
CREATE PROCEDURE/DROP PROCEDURE,CALL,GRANT/REVOKE EXECUTE,SECURITY INVOKER(default) andSECURITY DEFINER- Invoke from SQL and PGWire (
CALL), HTTP (POST /v1/functions/{namespace}/{procedure}), or a topic trigger - Nested
ctx.functions.call, transactionalctx.db.query/execute, stagedctx.topics.publish(commit flushes, rollback drops) - Inline JavaScript bodies, or project TypeScript compiled by
kalam deploy/kalam schema gen - Topic payload types:
CREATE TRIGGER … ON TOPIC … EXECUTE PROCEDUREwith a typedPAYLOAD CREATE SCHEDULE/ALTER SCHEDULE … ENABLE|DISABLE/DROP SCHEDULE(CRONorINTERVAL)- Catalog in
information_schema/pg_catalog.pg_proc; Admin UI for procedures, logs, and schedules - Generated TypeScript/Dart clients from the local schema (no live server required to compile the contract)
CREATE OR REPLACE PROCEDURE app.inc(x INT)
LANGUAGE JAVASCRIPT
AS $$
return input.x + 1;
$$;
CALL app.inc(41);Database types (#380)
Named types are the same contract for table columns and procedure signatures.
CREATE TYPE … AS (…)composites,AS ENUM, andFROM TABLErow-type aliases- Implicit row type per table (
app.usersis both the table and the row type); topics catalog an implicit payload type - Nested
STRUCT/Listpersist without JSON fallbacks ALTER TYPE(attributes and enum values),DROP TYPE,COMMENT ON TYPE/COMMENT ON PROCEDURESET search_path(PostgreSQL spelling;USEremains an alias)
CREATE TYPE app.address AS (
city TEXT NOT NULL,
country TEXT NOT NULL
);
CREATE SHARED TABLE app.users (
id TEXT PRIMARY KEY,
name TEXT NOT NULL,
home app.address
);
CREATE TYPE app.user FROM TABLE app.users;Indexes on USER and SHARED tables (#380)
Scalar secondary indexes are equality prefix scans — not vector USING COSINE. They apply to USER and SHARED tables, including RLS membership lookups such as user_id.
CREATE INDEX/CREATE UNIQUE INDEX/DROP INDEX(alsoALTER TABLE … CREATE|DROP INDEX)- Hot prefix seek; cold segments prune with bloom / min-max instead of scanning every flushed file
- Round-trip through
kalam schema/ schema-diff (CREATE TYPE,CREATE INDEX,CREATE PROCEDURE)
CREATE INDEX idx_messages_conversation ON app.messages (conversation_id);
CREATE INDEX idx_conversation_members_user ON app.conversation_members (user_id);Everyday CLI (#380)
kalam now covers local databases, schema, and procedures — not only the SQL shell.
| Command | Purpose |
|---|---|
kalam init |
New app or template |
kalam dev |
Watch schema + procedures; start/status/logs/stop a background session |
kalam up / down |
Start / stop a local database (data kept on down) |
kalam status |
This folder, --global, or a named --env |
kalam instances |
Local servers and saved cloud connections (servers / list aliases) |
kalam logs |
Server logs (-F follow); --local-file for the capture file |
kalam schema gen |
Generate SDK types from local SQL |
kalam deploy |
Apply migrations and activate procedures (--dry-run to validate) |
kalam db migrate / reset / seed |
Migrations and fixtures |
kalam functions build / status / revisions / rollback / logs |
Procedure lifecycle |
--global / -g is the shared local database. --env selects a named project environment. --agent emits deterministic events for coding agents.
Also in this beta
- Direct PostgreSQL wire: JDBC/DBeaver-style
pg_catalogshims,search_path, procedure metadata, and typedCALL(#376, #380) - Admin UI: Functions, SQL Studio, Schedules; platform release binaries embed the UI
- Topic consumer groups: same-group fetch no longer overlaps, skips, or redelivers in-flight / already-acked offsets
- OIDC: JWKS rediscovery when Dex/IdP keys rotate, so invite login does not fail as “unauthorized”
- V8 isolate reuse, worker affinity, and idle memory accounting (#381)
- Windows MSVC link for V8 + RocksDB;
kalam updatecan refresh CLI and managed server
Not in this beta
- In-place upgrade of existing 0.6 data directories
- Table
AFTER ROWtriggers (route writes through a topic instead) - Extra function runtimes (V8 only)
CREATE TYPE … AS UNION/INTERFACE(reserved)- Vector indexes answering scalar equality
See the SQL reference and CLI guide.
What's Changed
- pgwire JDBC catalog, SQL Studio RLS policies, and CLI test-runner fixes by @jamals86 in #376
- Optimize V8 function session memory and warm reuse by @jamals86 in #381
- feat: 0.7 serialization, scalar indexes, and server functions by @jamals86 in #380
Full Changelog: v0.6.0-rc.0...v0.7.0-beta.0
What's Changed
- pgwire JDBC catalog, SQL Studio RLS policies, and CLI test-runner fixes by @jamals86 in #376
- Optimize V8 function session memory and warm reuse by @jamals86 in #381
- feat: 0.7 serialization, scalar indexes, and server functions by @jamals86 in #380
Full Changelog: v0.6.0-rc.0...v0.7.0-beta.0