memory-vault
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| VAULT_DIR | No | Data directory for memory vault | ~/.memory-vault |
| VAULT_TOP_K | No | Default number of retrieval results | 8 |
| VAULT_CONFIG | No | Path to configuration file | <data-dir>/config.json |
| VAULT_WEB_HOST | No | Web interface listen address | 127.0.0.1 |
| VAULT_WEB_PORT | No | Web interface listen port | 8988 |
| VAULT_KW_WEIGHT | No | Keyword weight in fusion ranking | 0.4 |
| VAULT_WEB_TOKEN | No | Bearer token for web API access; if set, all /api/* requests must include Authorization: Bearer <token> | |
| VAULT_DEDUP_MODE | No | Dedup behavior: merge or skip | merge |
| VAULT_EMBED_DIMS | No | Local embedding dimensions | 128 |
| VAULT_SEM_WEIGHT | No | Semantic weight in fusion ranking | 0.6 |
| VAULT_EMBED_MODEL | No | Embedding model name (depends on provider) | all-MiniLM-L6-v2 |
| VAULT_MIN_CLUSTER | No | Minimum cluster size | 2 |
| VAULT_RECENCY_DAYS | No | Time-decay half-life in days, 0 disables | 180 |
| VAULT_EMBED_API_KEY | No | API key for embedding service; supports env://VAR or file:///path forms | |
| VAULT_EMBED_API_URL | No | OpenAI-compatible API endpoint URL | |
| VAULT_EMBED_PROVIDER | No | Embedding provider: local, sentence, or api | local |
| VAULT_DEDUP_THRESHOLD | No | Write-time dedup similarity threshold | 0.92 |
| VAULT_EMBED_API_MODEL | No | Model name for API embedding provider | text-embedding-3-small |
| VAULT_CLUSTER_THRESHOLD | No | Clustering threshold for tidy | 0.78 |
| VAULT_DIGEST_MEMBER_CHARS | No | Characters retained per member record in summary | 600 |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| vault_putA | 存入一条记忆记录；自动去重（相同或高度相似的内容并入已有记录）。 |
| vault_askA | 混合检索记忆：关键词（BM25）与语义向量融合排序，返回按相关度降序的记录。 |
| vault_takeA | 按 id 查看单条记录的完整内容。 |
| vault_listA | 列出最近的记录（不含已折叠内容）。 |
| vault_dropC | 删除一条记录。 |
| vault_tidyA | 压缩记忆：把互相相似的记录折叠为一条摘要，原记录标记为已折叠。 |
| vault_statsA | 查看存储统计（记录数、向量后端、嵌入配置等）。 |
| vault_ingestB | 从 markdown 文件或目录批量导入记忆。 |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 8 tools
Each tool targets a distinct action: put for storing, ask for searching, take for reading by id, list for enumeration, drop for deletion, tidy for compaction, stats for metadata, and ingest for batch import. There is no overlap or ambiguity between these operations.
All tools follow a consistent vault_verb pattern using snake_case. Although 'stats' is a noun, it behaves as a verb in context and the pattern remains predictable and uniform.
Eight tools is well within the optimal 3-15 range and each serves a meaningful purpose for a memory vault. The count is neither excessive nor sparse for the domain.
The tool set covers core memory lifecycle: create/update via vault_put with deduplication, read via take/list/ask, delete via drop, plus organization via tidy and import via ingest. A minor gap is the lack of an explicit full-record edit, but the deduplication-based put mitigates this.