chore: remove eks deployment steps from ci workflow

[hyacinthus]

Remove unused EKS deployment steps from the CI build workflow.

[Copilot]

Pull request overview

This pull request removes unused EKS deployment steps from the CI workflow while also introducing checkpoint cleanup functionality and updating several Python dependencies.

Key Changes:

• Removes commented-out AWS EKS deployment steps from the GitHub Actions workflow
• Adds new checkpoint cleanup functionality with both a scheduled job and manual script
• Updates boto3 (1.41.4 → 1.41.5), botocore (1.41.4 → 1.41.5), langsmith (0.4.48 → 0.4.49), and rich-toolkit (0.16.0 → 0.17.0)

Reviewed changes

Copilot reviewed 4 out of 5 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
.github/workflows/build.yml	Removes unused AWS EKS deployment configuration and credentials setup (45 lines removed)
intentkit/core/cleanup.py	New module implementing checkpoint cleanup functionality with SQL-based deletion of old checkpoint data
scripts/cleanup_checkpoints.py	New CLI script for manual checkpoint cleanup with dry-run and force modes
intentkit/core/scheduler.py	Adds scheduled job to run checkpoint cleanup daily at 2:20 UTC
uv.lock	Routine dependency updates for boto3, botocore, langsmith, and rich-toolkit packages

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The SQL deletion logic may not correctly identify all old checkpoints. The query selects thread_id, checkpoint_ns from checkpoints where timestamp is old, but this may only match some checkpoints for a given thread_id. If a thread has both old and new checkpoints, this query will only delete the old ones, potentially leaving orphaned related records or creating an inconsistent state.

Consider whether the intent is to delete entire threads (all checkpoints for a thread_id where ANY checkpoint is old) or only specific old checkpoints. If it's the former, the logic should be adjusted to identify threads with old checkpoints, then delete ALL checkpoints for those threads.

Suggested change

	WITH old_threads AS (
	SELECT thread_id, checkpoint_ns
	FROM checkpoints
	WHERE (checkpoint ->> 'ts')::timestamptz < :cutoff
	),
	deleted_writes AS (
	DELETE FROM checkpoint_writes cw
	USING old_threads ot
	WHERE cw.thread_id = ot.thread_id AND cw.checkpoint_ns = ot.checkpoint_ns
	),
	deleted_blobs AS (
	DELETE FROM checkpoint_blobs cb
	USING old_threads ot
	WHERE cb.thread_id = ot.thread_id AND cb.checkpoint_ns = ot.checkpoint_ns
	)
	DELETE FROM checkpoints cp
	USING old_threads ot
	WHERE cp.thread_id = ot.thread_id AND cp.checkpoint_ns = ot.checkpoint_ns
	WITH threads_to_delete AS (
	SELECT DISTINCT thread_id
	FROM checkpoints
	WHERE (checkpoint ->> 'ts')::timestamptz < :cutoff
	),
	deleted_writes AS (
	DELETE FROM checkpoint_writes cw
	USING threads_to_delete ttd
	WHERE cw.thread_id = ttd.thread_id
	),
	deleted_blobs AS (
	DELETE FROM checkpoint_blobs cb
	USING threads_to_delete ttd
	WHERE cb.thread_id = ttd.thread_id
	)
	DELETE FROM checkpoints cp
	USING threads_to_delete ttd
	WHERE cp.thread_id = ttd.thread_id

[Copilot]

Mismatch between count and deletion logic. The count query (line 33-37) counts DISTINCT thread_id, suggesting it counts threads, but the deletion query deletes individual checkpoints matching the timestamp criteria. This means:

1. The returned count represents the number of threads with old checkpoints, not the number of checkpoints being deleted
2. The confirmation message in the script says "delete {count} threads and their history" but may only delete some checkpoints per thread
3. The log message "Found {thread_count} threads to delete" is misleading if not all checkpoints for those threads are deleted

Either rename thread_count to clarify what it represents, or adjust the deletion logic to match the counting logic.

[Copilot]

Missing test coverage for the new cleanup_checkpoints function. The function performs critical database operations (deletion of checkpoint data) and should have tests covering:

• Dry run mode behavior
• Actual deletion behavior
• Count accuracy
• Edge cases (no checkpoints to delete, boundary dates)
• Transaction rollback on errors

Similar functions in the codebase (e.g., tests/core/test_agent_asset.py, tests/models/test_user.py) have corresponding test files.

[Copilot]

The confirmation message is misleading. It says "delete {count} threads and their history" but based on the cleanup logic, it may only delete some checkpoints per thread, not entire threads. This inconsistency could lead to user confusion about what will actually be deleted.

The message should accurately reflect what the cleanup operation does, or the cleanup logic should be adjusted to match the message.

Suggested change

	f"Are you sure you want to delete {count} threads and their history? (y/N): "
	f"Are you sure you want to delete {count} checkpoint{'s' if count != 1 else ''}? (y/N): "

[Copilot]

Duplicate comment detected. The comment "Cleanup old checkpoints daily at UTC 2:20" appears twice in consecutive lines.

Suggested change

# Cleanup old checkpoints daily at UTC 2:20