Dimension 3: Super-Memory-TS Deep Analysis¶
Overview¶
- Repo: https://github.com/Veedubin/Super-Memory-TS
- NPM: @veedubin/super-memory-ts
- License: MIT
- Version: v2.6.5 (actively developed, 74 commits)
- By: Veedubin / Frozen Maple Labs
- Language: TypeScript (96.4%)
Core Architecture¶
- Type: MCP (Model Context Protocol) server
- Database: Qdrant (HNSW indexing, payload filtering)
- Embeddings: BGE-Large (GPU, 1024-dim) / MiniLM-L6-v2 (CPU, 384-dim)
- Precision: fp16 (~325MB), fp32 (~650MB), q8 (~162MB), q4 (~81MB)
- Vector Search: HNSW, sub-10ms query latency
- Local-first: All data stays on your machine
Key Capabilities¶
1. Semantic Memory Search¶
Store and retrieve memories using natural language queries with embedding-based similarity search.
2. Project Indexing¶
Automatically index project files for code-aware search: - Supports TypeScript, JavaScript, Python, Markdown, JSON, and more - Semantic chunking at function/class boundaries - Incremental updates via SHA-256 hash comparison - File watching with 500ms debounce (chokidar)
3. Project Isolation¶
Multi-project memory isolation using projectId tagging in Qdrant payloads. Set via BOOMERANG_PROJECT_ID env var.
4. Tiered Search Strategies¶
| Mode | Strategy | Best For |
|---|---|---|
| TIERED (default) | MiniLM primary + BGE fallback | Interactive queries, real-time responses |
| PARALLEL | Dual-tier with RRF fusion | Comprehensive recall, research tasks |
| vector_only | Pure semantic similarity | Exact semantic matching |
| text_only | Keyword matching via Fuse.js | Exact keyword matching |
5. MCP Tool Interface (5 tools)¶
| Tool | Purpose |
|---|---|
query_memories | Semantic search over stored memories |
add_memory | Store a new memory entry |
search_project | Search indexed project files |
index_project | Trigger project indexing |
get_file_contents | Reconstruct file contents from indexed chunks |
6. Multi-Collection Search with RRF¶
Search across multiple Qdrant collections (e.g., during model migration) with Reciprocal Rank Fusion.
7. Dual Integration Modes¶
- MCP Server (External): Standalone MCP server via HTTP/stdio
- Built-in (Boomerang): Core modules imported directly into Boomerang plugin
Memory Schema¶
interface MemoryEntry {
id: string; // UUID
text: string; // Content
vector: Float32Array; // 1024-dim embedding
sourceType: MemorySourceType; // session, file, web, boomerang, project
sourcePath?: string;
timestamp: Date;
contentHash: string; // SHA-256 for deduplication
metadataJson?: string;
}
Performance Benchmarks¶
- Semantic query (HNSW): <10ms p50
- Embedding generation: <100ms (BGE-Large single text)
- Batch embedding: <50ms/text (batch of 8)
- Memory add: ~20/sec
- Memory query: ~100/sec
- Project file indexing: ~100 files/min
Companion: Boomerang-v2¶
- Package: @veedubin/boomerang-v2
- Type: Multi-agent orchestration plugin for OpenCode
- Features: 14 specialized agents, 8-step protocol, built-in semantic memory
What Super-Memory-TS LACKS (compared to Hermes providers)¶
- No automatic memory extraction — memories must be explicitly added via
add_memory - No knowledge graph — pure vector similarity, no entity relationships
- No trust scoring — all memories treated equally
- No session-end extraction — no automatic summarization
- No dialectic reasoning — no LLM-synthesized insights about the user
- No tiered context loading — doesn't do L0/L1/L2 abstraction
- No cross-session user modeling — no persistent user profile
- No pre-compression extraction — doesn't capture before context compression
- No context fencing — no protection against recursive memory pollution
- No built-in memory types — no distinction between session/user/organizational memory
- No contradiction detection — doesn't detect conflicting memories
- No memory decay/consolidation — old memories never fade or get compressed