feat: Add conversation variable persistence layer

[hila-f-qodo]

Benchmark PR from qodo-benchmark#429

---

[devin-ai-integration]

Devin Review found 1 potential issue.

View issue and 5 additional flags in Devin Review.

[devin-ai-integration]

🔴 SQLAlchemy Session not closed in ConversationVariableUpdaterImpl.update()

The update method creates a SQLAlchemy Session without using a context manager, causing the session to never be closed. This leads to database connection leaks.

Click to expand

Problem

The original implementation (before the refactoring) used a context manager:

with Session(db.engine) as session:
    row = session.scalar(stmt)
    ...
    session.commit()

But the new implementation at api/services/conversation_variable_updater.py:18-23 does not:

session = Session(db.engine)
row = session.scalar(stmt)
if not row:
    raise ConversationVariableNotFoundError(...)
row.data = variable.model_dump_json()
session.commit()

Impact

• Database connections are never returned to the pool
• Over time, this can exhaust the database connection pool
• This method is called for every conversation variable update in the ConversationVariablePersistenceLayer.on_event() method (api/core/app/layers/conversation_variable_persist_layer.py:52)
• The AGENTS.md file explicitly states: "SQLAlchemy Sessions Must Use Context Managers" and "Database sessions are created without context managers or sessions are not properly closed" is a failure criteria

Recommendation: Use a context manager for the session:

def update(self, conversation_id: str, variable: Variable) -> None:
    stmt = select(ConversationVariable).where(
        ConversationVariable.id == variable.id, ConversationVariable.conversation_id == conversation_id
    )
    with Session(db.engine) as session:
        row = session.scalar(stmt)
        if not row:
            raise ConversationVariableNotFoundError("conversation variable not found in the database")
        row.data = variable.model_dump_json()
        session.commit()

---

Was this helpful? React with 👍 or 👎 to provide feedback.