refactor: remove AI slop

[sulthonzh]

refactor: remove AI slop

Removes unused imports (dead code) across the pipeline-service. Pure deletions of imports that have no references anywhere in the codebase — no logic, behavior, or public API changes.

Scope

Behavior-preserving cleanup only. No dependency, workflow, lockfile, or test changes.

Changes

Removed unused imports detected via AST scan and verified by grep:

File	Removed
pipeline-service/config/settings.py	from typing import Optional, import os
pipeline-service/database.py	from models.customer import Customer inside init_db() (unused local import)
pipeline-service/main.py	Optional from typing import, from models.customer import Customer
pipeline-service/services/ingestion.py	List, Optional from typing import
pipeline-service/services/pipeline.py	List, Optional from typing import
pipeline-service/sinks/postgres_sink.py	from database import get_db, from datetime import datetime, timezone
pipeline-service/sources/base.py	Optional from typing import

What was preserved

• __init__.py re-exports — intentional public API surface (consumers import from package roots).
• Public methods on repository/source/service classes (e.g. bulk_create, get_or_create, fetch_paginated, get_pipeline_status) — these are part of the framework's public API surface, not dead code.
• except Exception blocks in validate() methods on HTTPSource / PostgresSink and the orchestrator's execute() — these are I/O / top-level boundaries that intentionally convert any failure to a boolean / IngestResponse.
• Inline import statements inside endpoint handlers — lazy-loading pattern, not slop.

Verification

• Baseline (before): pytest tests/test_unit/ tests/test_integration/ → 28 passed, 0 failed
• After: pytest tests/test_unit/ tests/test_integration/ → 28 passed, 0 failed
• No new failures introduced. E2E suite requires Docker stack and is unchanged.

[github-actions]

AI Code Review ✅

No issues found. LGTM!