fix: worker waits for schema before claiming; npm ci for reproducible builds

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

worker/lyra_worker/db.py

@@ -1,4 +1,5 @@
 import os
+import time
 
 import psycopg
 
@@ -6,3 +7,34 @@ import psycopg
 def connect() -> psycopg.Connection:
     dsn = os.environ["DATABASE_URL"]
     return psycopg.connect(dsn)
+
+
+def wait_for_db(timeout: float = 60.0, interval: float = 2.0) -> psycopg.Connection:
+    """Block until the database is up and the `Job` table exists.
+
+    Repeatedly opens a connection and probes for the `Job` table's presence.
+    Tolerates the DB not being reachable yet (`psycopg.OperationalError`) and
+    migrations not having run yet (`psycopg.errors.UndefinedTable`), retrying
+    with a fixed `interval` until `timeout` elapses. Returns the open, usable
+    connection on success; re-raises the last error on timeout.
+    """
+    deadline = time.monotonic() + timeout
+    printed_waiting = False
+    last_error: Exception | None = None
+    while time.monotonic() < deadline:
+        conn: psycopg.Connection | None = None
+        try:
+            conn = connect()
+            with conn.cursor() as cur:
+                cur.execute('SELECT 1 FROM "Job" LIMIT 0')
+            return conn
+        except (psycopg.OperationalError, psycopg.errors.UndefinedTable) as exc:
+            last_error = exc
+            if conn is not None:
+                conn.close()
+            if not printed_waiting:
+                print("worker: waiting for database schema...", flush=True)
+                printed_waiting = True
+            time.sleep(interval)
+    assert last_error is not None
+    raise last_error

worker/lyra_worker/main.py

@@ -1,7 +1,7 @@
 import time
 
 from lyra_worker.claim import claim_next
-from lyra_worker.db import connect
+from lyra_worker.db import wait_for_db
 from lyra_worker.pipeline import run_pipeline
 
 IDLE_SLEEP = 2.0
@@ -9,7 +9,7 @@ STAGE_DELAY = 1.0  # visible progress in the UI
 
 
 def run_forever() -> None:
-    conn = connect()
+    conn = wait_for_db()
     print("worker: waiting for jobs", flush=True)
     try:
         while True:

worker/tests/test_db.py

@@ -0,0 +1,11 @@
+from lyra_worker.db import wait_for_db
+
+
+def test_wait_for_db_returns_usable_connection_when_schema_ready(conn):
+    # `conn` fixture guarantees the DB is up and the schema is migrated.
+    result = wait_for_db(timeout=5)
+    try:
+        with result.cursor() as cur:
+            cur.execute('SELECT 1 FROM "Job" LIMIT 0')
+    finally:
+        result.close()