me_memory_search

Search and browse memories using text matching and/or filters.

Supports three search modes: semantic (meaning-based), fulltext (keyword-based via BM25), or hybrid (both combined via Reciprocal Rank Fusion). Combine any search mode with tree, meta, and temporal filters.

Parameters

Name	Type	Required	Description
semantic	string \| null	yes	Natural language query for semantic search. Pass null to skip.
fulltext	string \| null	yes	Keywords/phrases for BM25 exact matching. Pass null to skip.
grep	string \| null	yes	POSIX regex pattern filter on content (case-insensitive). Applied as a WHERE filter alongside other filters. Pass null to skip.
meta	object \| null	yes	Filter by metadata attributes. Pass null to skip.
tree	string \| null	yes	Filter by tree path. Pass null to skip.
temporal	object \| null	yes	Temporal filter. Pass null to skip.
weights	object \| null	yes	Weights for hybrid search ranking. Pass null for defaults.
candidateLimit	integer	yes	Candidates per search mode before RRF fusion. Pass 0 for default (30).
limit	integer	yes	Maximum number of results. Pass 0 for default (10). Max: 1000.
order_by	string \| null	yes	Sort direction for filter-only searches: "asc" or "desc". Default: "desc". Pass null for default.

tree syntax

• Bare path (work.projects) -- matches that exact node only.
• Wildcard (work.projects.*) -- matches all descendants.
• lquery pattern (*.api.*) -- matches any path containing api.
• ltxtquery (api & v2) -- label search with boolean operators.

temporal

Name	Type	Required	Description
contains	string \| null	yes	Find memories whose time range contains this point in time.
overlaps	object \| null	yes	Find memories overlapping this range ({start, end}).
within	object \| null	yes	Find memories fully within this range ({start, end}).

weights

Name	Type	Required	Description
fulltext	number \| null	yes	Weight for BM25 keyword matching (0-1).
semantic	number \| null	yes	Weight for semantic similarity (0-1).

Returns

{
  "results": [
    {
      "id": "0194a000-0001-7000-8000-000000000001",
      "content": "Use ltree for hierarchical path queries.",
      "meta": { "source": "docs" },
      "tree": "research.postgres",
      "temporal": null,
      "hasEmbedding": true,
      "createdAt": "2025-04-15T12:00:00Z",
      "createdBy": "user_abc",
      "updatedAt": null,
      "score": 0.85
    }
  ],
  "total": 1,
  "limit": 10
}

Field	Type	Description
results	array	Array of memory objects, each with an additional score field (0-1).
total	number	Total number of matching memories.
limit	number	The limit that was applied.

Examples

Semantic search

{
  "semantic": "how does authentication work",
  "fulltext": null,
  "grep": null,
  "meta": null,
  "tree": null,
  "temporal": null,
  "weights": null,
  "candidateLimit": 0,
  "limit": 10,
  "order_by": null
}

Hybrid search with tree filter

{
  "semantic": "embedding performance",
  "fulltext": "nomic ollama",
  "grep": null,
  "meta": null,
  "tree": "me.design.*",
  "temporal": null,
  "weights": null,
  "candidateLimit": 0,
  "limit": 5,
  "order_by": null
}

Filter-only browse (no search)

{
  "semantic": null,
  "fulltext": null,
  "grep": null,
  "meta": { "type": "decision" },
  "tree": "me.strategy.*",
  "temporal": null,
  "weights": null,
  "candidateLimit": 0,
  "limit": 20,
  "order_by": "desc"
}

Notes

• Provide at least one of semantic, fulltext, or a filter (tree, meta, temporal, grep) -- otherwise the search has no criteria.
• When both semantic and fulltext are provided, results are ranked using Reciprocal Rank Fusion (hybrid mode).
• order_by only applies to filter-only searches (no semantic/fulltext). Ranked searches are always sorted by score.
• score ranges from 0 to 1, where 1 is the best match.