AI-Driven SQL Performance Impact Analysis in CI/CD

Act as an AI-Assisted Database Architect. Design a revolutionary CI/CD pipeline step that weaponizes a highly capable [LLM_MODEL] (e.g., GPT-4o or Claude 3.5 Sonnet) to autonomously analyze the true performance impact of a proposed database migration. The Python automation script must intercept the raw SQL from a GitHub Pull Request, execute it against a mirrored staging [DATABASE_ENGINE] instance, and extract the raw, highly complex `EXPLAIN ANALYZE` query plan output. It must then feed this [EXPLAIN_ANALYZE_OUTPUT] directly into the LLM via API. The AI must be prompted to aggressively flag catastrophic performance regressions: identifying massive sequential/full table scans, severe memory sort usage, or glaringly missing indexes. The script must parse the AI's response and automatically post a 'Reviewer' comment directly on the PR, physically blocking the code merge if the projected 'Cost' integer exceeds a hardcoded threshold.