LTM Cache Management & Repair
Protocol for auditing, repairing, or purging problematic entries within the Long-Term Memory (LTM) cache to maintain data integrity.
Prompt Example
Use the skill 'ltm-cache-cleanup' to inspect and repair the following issue in the LTM cache: [PROBLEM/ERROR MESSAGE].
Concrete examples
Use the skill 'ltm-cache-cleanup' to inspect and repair the following issue in the LTM cache: The entry 'academy-skill-redesign-decisions' shows a retrieval error in semantic search.
Maintain the Long-Term Memory (LTM) environment by addressing data corruption, failed writes, or retrieval inconsistencies within the cache.
The workflow involves auditing failed and non-retryable entries to determine the appropriate course of action: requeueing (via `requeue_ltm_cache_entry`), permanent deletion (via `delete_ltm_cache_entry`), or executing a replacement episode with corrected semantic content.
For cases involving low signal extraction, facts must be refactored into high-fidelity statements prior to reintegration. In the event of ambiguous errors, thorough diagnostics are mandatory before attempting any retry operations.
The objective is to ensure a high-signal LTM state, preventing the accumulation of stale or corrupted records from degrading the reliability of the memory graph.