Session Logging Change Log¶

Date: 2025-08-18T12:21:47.014117+00:00

Summary¶

Session logging was introduced to give AI agents a durable, queryable audit trail of every coding session. All events are written to a SQLite database, making it easy to replay, search, and analyse agent behaviour over time.

Changes¶

Added src/codex/logging/session_logger.py (SQLite-backed).
Patched entrypoint.sh with start/end logging hooks (idempotent).
Created tests/test_session_logging.py (pytest).
Appended README with usage and schema.
Guardrail: No changes to .github/workflows and no activation of Actions.

Usage¶

from codex.logging.session_logger import SessionLogger

logger = SessionLogger()
logger.log(role="assistant", content="Session started")

Query logs via:

python -m codex.logging.query_logs --session-id <id>

Schema¶

Column	Type	Description
`id`	INTEGER	Auto-increment primary key
`session_id`	TEXT	UUID grouping all events in one session
`role`	TEXT	`system`, `user`, `assistant`, or `tool`
`content`	TEXT	Event payload
`ts`	TEXT	ISO-8601 timestamp