Hybrid retrieval
Three signals, one query.
A single query spans graph links, semantic vector matches, and wiki context. The candidates merge, and on paid plans a premium reranker sharpens which memories surface first.
One query fans out to graph, vector, and wiki signals, then merges and reranks them into a single ordered result.
Graph + vector + wiki
Every signal searched in one pass
Most memory tools make you pick a lane: a vector store, or a graph, or full-text. Hybrid retrieval searches all three at once. Graph links, semantic vector matches, and long-form wiki context are queried together, so a single ask reaches everything your agents know.
- Graph + vector + wiki — every signal searched in one pass
- Merged, not siloed — results come back as one ranked set
- One ask, full context — no separate calls to stitch together
Premium reranker
An extra ranking pass on Pro and above
Merging is only half the story: order matters. On paid plans, a premium reranker takes a second pass over the merged candidates to sharpen which memories surface first, so the most relevant context lands at the top of the result.
- Premium reranker — an extra ranking pass on Pro and above
- Sharper top results — the most relevant memories surface first
Keep exploring