Querying Memories¶

Lerim provides several ways to search and retrieve memories. All queries are read-only and project-scoped.

`lerim ask` -- LLM-powered Q&A¶

The primary query interface. Sends your question to the lead agent with memory context.

Requires running server

lerim ask is a service command that requires lerim up or lerim serve to be running.

Basic query¶

lerim ask "Why did we choose Postgres over SQLite?"

The lead agent retrieves relevant memories, uses them as context, and returns a natural language answer with evidence of which memories were consulted.

Limit context¶

Control how many memory items are included as context:

lerim ask "What auth pattern do we use?" --limit 5

Flag	Default	Description
`question`	required	Your question (quote if it contains spaces)
`--project`	--	Scope to a specific project (not yet implemented)
`--limit`	`12`	Max memory items provided as context

JSON output¶

Get structured output for scripting or agent integration:

lerim ask "How is the database configured?" --json

Returns JSON with the answer, sources, and metadata.

`lerim memory search` -- keyword search¶

Full-text keyword search across memory titles, bodies, and tags. Runs locally on the host -- no server required.

lerim memory search "database migration"

lerim memory search pytest --limit 5

Flag	Default	Description
`query`	required	Search string to match (case-insensitive)
`--project`	--	Filter to project (not yet implemented)
`--limit`	`20`	Max results

When to use search vs ask

Use memory search for quick keyword lookups when you know what you're looking for. Use ask when you need the LLM to reason about your question and synthesize an answer from multiple memories.

`lerim memory list` -- browse all memories¶

List stored memories (decisions and learnings), ordered by recency:

lerim memory list

lerim memory list --limit 10

lerim memory list --json

Flag	Default	Description
`--project`	--	Filter to project (not yet implemented)
`--limit`	`50`	Max items

Tips for effective queries¶

Be specific¶

# Good -- specific topic
lerim ask "What authentication pattern does the API use?"

# Less effective -- too broad
lerim ask "How does auth work?"

Reference past decisions¶

lerim ask "Why did we switch from REST to gRPC for the internal API?"
lerim ask "What problems did we have with the original caching approach?"

Check before implementing¶

At the start of a coding session, ask your agent:

Check lerim for any relevant memories about [topic you're working on].

Your agent will run lerim ask or lerim memory search to pull in past decisions and learnings before it starts working.

Combine search and ask¶

# Quick lookup to see what exists
lerim memory search "database"

# Then ask a specific question
lerim ask "What was the rationale for the database migration strategy?"