Obsidian MCP Server

# Search Result Cache LRU cache for search query results with TTL expiration. ## Purpose The Search Result Cache stores the results of search queries to avoid re-executing expensive search operations for repeated queries. ## How It Works ```mermaid flowchart TD A[Search request] --> B[Generate cache key] B --> C{In memory cache?} C -->|Yes| D{TTL expired?} C -->|No| E{In persistent cache?} D -->|No| F[Return cached results] D -->|Yes| G[Invalidate entry] E -->|Yes| H{TTL expired?} E -->|No| I[Execute search] H -->|No| J[Promote to memory] H -->|Yes| G G --> I I --> K{Large result?} K -->|Yes| L{Compression enabled?} K -->|No| M[Store in cache] L -->|Yes| N[Compress & store] L -->|No| M M --> F N --> F J --> F ``` ## Configuration | Variable | Type | Default | Description | | --------------------------------- | --------- | ------- | -------------------------------------- | | `OBSIDIAN_SEARCH_CACHE_SIZE` | `number` | `50` | Maximum entries in cache | | `OBSIDIAN_SEARCH_CACHE_TTL` | `number` | `30000` | TTL in milliseconds (30 seconds) | | `OBSIDIAN_COMPRESS_LARGE_RESULTS` | `boolean` | `false` | Enable compression for large results | | `OBSIDIAN_COMPRESSION_THRESHOLD` | `number` | `10240` | Size threshold for compression (bytes) | ## Cache Key Generation Cache keys are generated from search parameters: ``` query={query}&glob={glob}&regex={regex}&tags={tags}&frontmatter={fm}&limit={limit}&offset={offset} ``` This ensures different search configurations get separate cache entries. ## TTL Expiration Unlike the content cache, search results have a time-to-live: - Results expire after `OBSIDIAN_SEARCH_CACHE_TTL` milliseconds - Expired entries are removed on next access - Short TTL ensures results reflect recent vault changes ## Compression When enabled, large search results are compressed: ```mermaid flowchart LR A[Search Results] --> B{Size > threshold?} B -->|Yes| C[zlib compress] B -->|No| D[Store raw] C --> E[Store compressed] ``` See [Search Compression](search-compression.md) for details. ## Invalidation The cache is invalidated when: 1. **TTL expires**: Entries older than TTL are refreshed 2. **File modified**: All search cache cleared 3. **File created**: All search cache cleared 4. **File deleted**: All search cache cleared Note: Search cache is cleared entirely on any file change because determining which queries are affected is complex. ## Statistics Available via [get_cache_stats](../tools/get_cache_stats.md): | Stat | Description | | ------------------ | ------------------------------ | | `size` | Current entries in memory | | `maxSize` | Maximum cache capacity | | `ttl` | Configured TTL in milliseconds | | `persistentSize` | Entries in LMDB (if enabled) | | `compressedCount` | Number of compressed entries | | `compressionRatio` | Average compression ratio | ## Performance Impact | Scenario | Without Cache | With Cache | | ------------------------ | ------------- | ---------- | | First search (100 files) | ~500ms | ~500ms | | Repeated search | ~500ms | <5ms | | Different offset | ~500ms | <5ms | ## Source Files - [`src/services/cache/SearchResultCache.ts`](../../src/services/cache/SearchResultCache.ts) ## Related Features - [Persistent Cache](persistent-cache.md) - LMDB backing - [Search Compression](search-compression.md) - Result compression - [Inverted Index](inverted-index.md) - Faster search execution