openclaw/docs/concepts/memory-search.md

summary, title, read_when

summary	title	read_when
How memory search finds relevant notes using embeddings and hybrid retrieval	Memory search	You want to understand how memory_search works You want to choose an embedding provider You want to tune search quality

memory_search finds relevant notes from your memory files, even when the wording differs from the original text. It works by indexing memory into small chunks and searching them using embeddings, keywords, or both.

Quick start

If you have a GitHub Copilot subscription, OpenAI, Gemini, Voyage, or Mistral API key configured, memory search works automatically. To set a provider explicitly:

{
  agents: {
    defaults: {
      memorySearch: {
        provider: "openai", // or "gemini", "local", "ollama", etc.
      },
    },
  },
}

For multi-endpoint setups, provider can also be a custom models.providers.<id> entry, such as ollama-5080, when that provider sets api: "ollama" or another embedding adapter owner.

For local embeddings with no API key, set provider: "local". Source checkouts may still require native build approval: pnpm approve-builds then pnpm rebuild node-llama-cpp.

Some OpenAI-compatible embedding endpoints require asymmetric labels such as input_type: "query" for searches and input_type: "document" or "passage" for indexed chunks. Configure those with memorySearch.queryInputType and memorySearch.documentInputType; see the Memory configuration reference.

Supported providers

Provider	ID	Needs API key	Notes
Bedrock	bedrock	No	Auto-detected when the AWS credential chain resolves
Gemini	gemini	Yes	Supports image/audio indexing
GitHub Copilot	github-copilot	No	Auto-detected, uses Copilot subscription
Local	local	No	GGUF model, ~0.6 GB download
Mistral	mistral	Yes	Auto-detected
Ollama	ollama	No	Local, must set explicitly
OpenAI	openai	Yes	Auto-detected, fast
Voyage	voyage	Yes	Auto-detected

How search works

OpenClaw runs two retrieval paths in parallel and merges the results:

flowchart LR
    Q["Query"] --> E["Embedding"]
    Q --> T["Tokenize"]
    E --> VS["Vector Search"]
    T --> BM["BM25 Search"]
    VS --> M["Weighted Merge"]
    BM --> M
    M --> R["Top Results"]

• Vector search finds notes with similar meaning ("gateway host" matches "the machine running OpenClaw").
• BM25 keyword search finds exact matches (IDs, error strings, config keys).

If only one path is available (no embeddings or no FTS), the other runs alone.

When embeddings are unavailable, OpenClaw still uses lexical ranking over FTS results instead of falling back to raw exact-match ordering only. That degraded mode boosts chunks with stronger query-term coverage and relevant file paths, which keeps recall useful even without sqlite-vec or an embedding provider.

Improving search quality

Two optional features help when you have a large note history:

Temporal decay

Old notes gradually lose ranking weight so recent information surfaces first. With the default half-life of 30 days, a note from last month scores at 50% of its original weight. Evergreen files like MEMORY.md are never decayed.

Enable temporal decay if your agent has months of daily notes and stale information keeps outranking recent context.

MMR (diversity)

Reduces redundant results. If five notes all mention the same router config, MMR ensures the top results cover different topics instead of repeating.

Enable MMR if `memory_search` keeps returning near-duplicate snippets from different daily notes.

Enable both

{
  agents: {
    defaults: {
      memorySearch: {
        query: {
          hybrid: {
            mmr: { enabled: true },
            temporalDecay: { enabled: true },
          },
        },
      },
    },
  },
}

Multimodal memory

With Gemini Embedding 2, you can index images and audio files alongside Markdown. Search queries remain text, but they match against visual and audio content. See the Memory configuration reference for setup.

Session memory search

You can optionally index session transcripts so memory_search can recall earlier conversations. This is opt-in via memorySearch.experimental.sessionMemory. See the configuration reference for details.

Troubleshooting

No results? Run openclaw memory status to check the index. If empty, run openclaw memory index --force.

Only keyword matches? Your embedding provider may not be configured. Check openclaw memory status --deep.

Local embeddings time out? ollama, lmstudio, and local use a longer inline batch timeout by default. If the host is simply slow, set agents.defaults.memorySearch.sync.embeddingBatchTimeoutSeconds and rerun openclaw memory index --force.

CJK text not found? Rebuild the FTS index with openclaw memory index --force.

Further reading

• Active Memory -- sub-agent memory for interactive chat sessions
• Memory -- file layout, backends, tools
• Memory configuration reference -- all config knobs

Related

• Memory overview
• Active memory
• Builtin memory engine