Promote dev to staging: topics infra ownership and secret fallback hardening

coaching/src/core/config_multitenant.py

@@ -235,15 +235,17 @@ def get_openai_api_key() -> str | None:
         return settings.openai_api_key
 
     # Retrieve from Secrets Manager.
-    # Prefer configured value when provided, and keep legacy fallback for compatibility.
-    # Otherwise, use stage-aware naming and then legacy global fallback.
-    secret_candidates = (
-        [settings.openai_api_key_secret, "purposepath/openai-api-key"]
-        if settings.openai_api_key_secret
-        else [
-            f"purposepath/{settings.stage}/openai-api-key",
-            "purposepath/openai-api-key",
-        ]
+    # Always try configured name first, then stage-aware default, then legacy global.
+    # This keeps compatibility when secrets are renamed between conventions.
+    stage_default = f"purposepath/{settings.stage}/openai-api-key"
+    secret_candidates = list(
+        dict.fromkeys(
+            [
+                settings.openai_api_key_secret,
+                stage_default,
+                "purposepath/openai-api-key",
+            ]
+        )
     )
     try:
         from shared.services.aws_helpers import get_secretsmanager_client
@@ -282,20 +284,17 @@ def get_google_vertex_credentials() -> dict[str, Any] | None:
 
     settings = get_settings()
 
-    # Always retrieve from Secrets Manager (ignore GOOGLE_APPLICATION_CREDENTIALS)
-    # This ensures proper OAuth scopes are applied via service_account.Credentials.
-    # Prefer configured value when provided, and keep legacy fallback for compatibility.
-    # Otherwise, use stage-aware naming and then legacy global fallback.
-    secret_candidates = (
-        [
-            settings.google_vertex_credentials_secret,
-            "purposepath/google-vertex-credentials",
-        ]
-        if settings.google_vertex_credentials_secret
-        else [
-            f"purposepath/{settings.stage}/google-vertex-credentials",
-            "purposepath/google-vertex-credentials",
-        ]
+    # Always retrieve from Secrets Manager (ignore GOOGLE_APPLICATION_CREDENTIALS).
+    # Try configured name, then stage-aware default, then legacy global fallback.
+    stage_default = f"purposepath/{settings.stage}/google-vertex-credentials"
+    secret_candidates = list(
+        dict.fromkeys(
+            [
+                settings.google_vertex_credentials_secret,
+                stage_default,
+                "purposepath/google-vertex-credentials",
+            ]
+        )
     )
     try:
         import structlog

infrastructure/pulumi/__main__.py

@@ -82,6 +82,27 @@
     tags={**common_tags, "Name": "coaching_sessions", "Purpose": "Session-Tracking"},
 )
 
+# Topic definitions table (master data consumed by coaching service runtime)
+topics_table = aws.dynamodb.Table(
+    "topics",
+    name=f"purposepath-topics-{stack}",
+    billing_mode="PAY_PER_REQUEST",
+    hash_key="topic_id",
+    attributes=[
+        aws.dynamodb.TableAttributeArgs(name="topic_id", type="S"),
+        aws.dynamodb.TableAttributeArgs(name="topic_type", type="S"),
+    ],
+    global_secondary_indexes=[
+        aws.dynamodb.TableGlobalSecondaryIndexArgs(
+            name="type-index",
+            hash_key="topic_type",
+            projection_type="ALL",
+        ),
+    ],
+    point_in_time_recovery=aws.dynamodb.TablePointInTimeRecoveryArgs(enabled=True),
+    tags={**common_tags, "Name": "topics", "Purpose": "Topic-Definitions"},
+)
+
 # S3 Bucket for LLM Prompts
 prompts_bucket = aws.s3.Bucket(
     "coaching-prompts-bucket",
@@ -153,13 +174,15 @@
     {
         "coachingConversations": conversations_table.name,
         "coachingSessions": coaching_sessions_table.name,
+        "topics": topics_table.name,
     },
 )
 pulumi.export(
     "tableNames",
     {
         "conversations": f"purposepath-coaching-conversations-{stack}",
         "sessions": f"purposepath-coaching-sessions-{stack}",
+        "topics": f"purposepath-topics-{stack}",
     },
 )
 pulumi.export("promptsBucket", prompts_bucket.bucket)
@@ -175,5 +198,6 @@
     {
         "conversations": conversations_table.arn,
         "sessions": coaching_sessions_table.arn,
+        "topics": topics_table.arn,
     },
 )