UnisonWhat's the difference between auto, standard, and deep recall modes?
Choose the right recall mode for your use case: auto for most situations, standard for speed, deep for thorough aggregation.
When should I use mode=auto vs mode=standard vs mode=deep?
The mode parameter on GET /v1/brain/context controls how aggressively the brain searches and aggregates before returning context.
The three modes
| Mode | Behavior | When to use |
|---|---|---|
standard | Fast single-pass vector search. Returns the top-k most similar chunks. | Low-latency paths where you need a quick answer and the topic is well-represented in the brain. |
deep | Multi-pass search with entity resolution and cross-source aggregation. Slower, but surfaces facts that are spread across many small ingested pieces. | Questions about people, decisions, or topics where evidence is distributed. Reduces weakEvidence responses. |
auto | Picks standard or deep based on query complexity and estimated brain coverage. | Default for most cases — let the brain decide. |
Example
# Fast lookup
GET /v1/brain/context?q=what+database+do+we+use&mode=standard
# Thorough aggregation
GET /v1/brain/context?q=everything+we+know+about+Alice&mode=deep
# Let the brain decide
GET /v1/brain/context?q=last+API+design+decision&mode=autoPractical guidance
Start with mode=auto. Switch to mode=deep if you're getting weakEvidence: true on topics you know are in the brain. Use mode=standard only if latency is critical and auto is measurably slower for your workload.
What does weakEvidence mean in a recall response?
Understand when Unison returns weakEvidence:true and what to do about it.
How do I improve the quality of what the brain returns?
Practical steps to get better recall from the Unison brain: ingest more, write clearer facts, use the right mode, and scope your queries.