Add instance table along with init code.

This change adds an `instance` table containing the minimum set of
details about codegate. We might want to add more details in the
future.

The table must contain only a single record at any time. To guarantee
that, a trigger is added that prevents inserts beyond the first
record.

Finally, code performing the initialization is added to the `serve`
command.

Author: blkt

migrations/versions/2025_03_05_2126-e4c05d7591a8_add_installation_table.py

@@ -0,0 +1,63 @@
+"""add installation table
+
+Revision ID: e4c05d7591a8
+Revises: 3ec2b4ab569c
+Create Date: 2025-03-05 21:26:19.034319+00:00
+
+"""
+
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision: str = "e4c05d7591a8"
+down_revision: Union[str, None] = "3ec2b4ab569c"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    op.execute("BEGIN TRANSACTION;")
+
+    op.execute(
+        """
+        CREATE TABLE IF NOT EXISTS instance (
+          id TEXT PRIMARY KEY,  -- UUID stored as TEXT
+          created_at DATETIME NOT NULL
+        );
+        """
+    )
+
+    op.execute(
+        """
+        -- The following trigger prevents multiple insertions in the
+        -- instance table. It is safe since the dimension of the table
+        -- is fixed.
+
+        CREATE TRIGGER single_instance
+        BEFORE INSERT ON instance
+        WHEN (SELECT COUNT(*) FROM instance) >= 1
+        BEGIN
+          SELECT RAISE(FAIL, 'only one instance!');
+        END;
+        """
+    )
+
+    # Finish transaction
+    op.execute("COMMIT;")
+
+
+def downgrade() -> None:
+    op.execute("BEGIN TRANSACTION;")
+
+    op.execute(
+        """
+        DROP TABLE instance;
+        """
+    )
+
+    # Finish transaction
+    op.execute("COMMIT;")

src/codegate/cli.py

@@ -14,7 +14,11 @@
 from codegate.ca.codegate_ca import CertificateAuthority
 from codegate.codegate_logging import LogFormat, LogLevel, setup_logging
 from codegate.config import Config, ConfigurationError
-from codegate.db.connection import init_db_sync, init_session_if_not_exists
+from codegate.db.connection import (
+    init_db_sync,
+    init_session_if_not_exists,
+    init_instance,
+)
 from codegate.pipeline.factory import PipelineFactory
 from codegate.pipeline.sensitive_data.manager import SensitiveDataManager
 from codegate.providers import crud as provendcrud
@@ -318,6 +322,7 @@ def serve(  # noqa: C901
         logger = structlog.get_logger("codegate").bind(origin="cli")
 
         init_db_sync(cfg.db_path)
+        init_instance(cfg.db_path)
         init_session_if_not_exists(cfg.db_path)
 
         # Check certificates and create CA if necessary

src/codegate/db/connection.py

@@ -1,4 +1,5 @@
 import asyncio
+import datetime
 import json
 import sqlite3
 import uuid
@@ -23,6 +24,7 @@
     Alert,
     GetPromptWithOutputsRow,
     GetWorkspaceByNameConditions,
+    Instance,
     IntermediatePromptWithOutputUsageAlerts,
     MuxRule,
     Output,
@@ -570,6 +572,27 @@ async def add_persona(self, persona: PersonaEmbedding) -> None:
             logger.debug(f"Exception type: {type(e)}")
             raise AlreadyExistsError(f"Persona '{persona.name}' already exists.")
 
+    async def init_instance(self) -> None:
+        """
+        Initializes instance details in the database.
+        """
+        sql = text(
+            """
+            INSERT INTO instance (id, created_at)
+            VALUES (:id, :created_at)
+            """
+        )
+
+        try:
+            instance = Instance(
+                id=str(uuid.uuid4()),
+                created_at=datetime.datetime.now(datetime.timezone.utc),
+            )
+            await self._execute_with_no_return(sql, instance.model_dump())
+        except IntegrityError as e:
+            logger.debug(f"Exception type: {type(e)}")
+            raise AlreadyExistsError(f"Instance already initialized.")
+
 
 class DbReader(DbCodeGate):
     def __init__(self, sqlite_path: Optional[str] = None, *args, **kwargs):
@@ -1045,6 +1068,13 @@ async def get_distance_to_persona(
         )
         return persona_distance[0]
 
+    async def get_instance(self) -> Instance:
+        """
+        Get the details of the instance.
+        """
+        sql = text("SELECT id, created_at FROM instance")
+        return await self._execute_select_pydantic_model(Instance, sql)
+
 
 class DbTransaction:
     def __init__(self):
@@ -1095,8 +1125,6 @@ def init_db_sync(db_path: Optional[str] = None):
 
 
 def init_session_if_not_exists(db_path: Optional[str] = None):
-    import datetime
-
     db_reader = DbReader(db_path)
     sessions = asyncio.run(db_reader.get_sessions())
     # If there are no sessions, create a new one
@@ -1116,5 +1144,19 @@ def init_session_if_not_exists(db_path: Optional[str] = None):
         logger.info("Session in DB initialized successfully.")
 
 
+def init_instance(db_path: Optional[str] = None):
+    db_reader = DbReader(db_path)
+    instance = asyncio.run(db_reader.get_instance())
+    # Initialize instance if not already initialized.
+    if not instance:
+        db_recorder = DbRecorder(db_path)
+        try:
+            asyncio.run(db_recorder.init_instance())
+        except Exception as e:
+            logger.error(f"Failed to initialize instance in DB: {e}")
+            raise
+        logger.info("Instance initialized successfully.")
+
+
 if __name__ == "__main__":
     init_db_sync()

src/codegate/db/models.py

@@ -120,6 +120,11 @@ class Session(BaseModel):
     last_update: datetime.datetime
 
 
+class Instance(BaseModel):
+    id: str
+    created_at: datetime.datetime
+
+
 # Models for select queries