# Valkey Advisory Cache Service

## Module
Concelier

## Status
IMPLEMENTED

## Description
Valkey (Redis-compatible) caching layer for canonical advisories with TTL policies based on interest score, PURL index lookups, hot set ranking, and p99 < 20ms read target. Includes cache warmup, metrics, and fallback mode.

## Implementation Details
- **Modules**: `src/Concelier/__Libraries/StellaOps.Concelier.Cache.Valkey/`
- **Key Classes**:
  - `ValkeyAdvisoryCacheService` (`src/Concelier/__Libraries/StellaOps.Concelier.Cache.Valkey/ValkeyAdvisoryCacheService.cs`) - Valkey-backed cache with TTL policies, PURL index, and hot set ranking
  - `AdvisoryCacheKeys` (`src/Concelier/__Libraries/StellaOps.Concelier.Cache.Valkey/AdvisoryCacheKeys.cs`) - deterministic cache key generation for advisory lookups
  - `CachingCanonicalAdvisoryService` (`src/Concelier/__Libraries/StellaOps.Concelier.Core/Canonical/CachingCanonicalAdvisoryService.cs`) - caching decorator wrapping the canonical advisory service
- **Interfaces**: `IAdvisoryCacheService`
- **Source**: SPRINT_8200_0013_0001_GW_valkey_advisory_cache.md

## E2E Test Plan
- [ ] Store a canonical advisory in Valkey and retrieve it, verifying p99 read latency is < 20ms
- [ ] Verify TTL policy: high-interest advisories have longer TTLs than low-interest ones
- [ ] Verify PURL index lookup: query by PURL and verify matching advisories are returned from cache
- [ ] Verify hot set ranking: frequently accessed advisories are promoted in the hot set
- [ ] Verify cache warmup: on startup, verify high-interest advisories are pre-loaded into cache
- [ ] Verify fallback mode: disable Valkey and confirm the service falls back to direct PostgreSQL queries