Capture

The best notes are the ones you never got around to writing. Capture is the pad writing them for you: what happened in a session becomes durable notes, without slowing anything down and without dumping raw noise into the pad.

The rest of this section is how that's kept honest.

A two-tier write. Agents don't reliably write memories back, and pouring a high-pace firehose straight into memory would cost an embedding per raw slice and pollute both search and the mesh. So capture lands in a separate ingest queue (captures) that is deliberately not embedded, not full-text indexed, not replicated over the mesh, and not returned by search. Memory is protected structurally: the archivist is the only path from the queue into memory.

Off the hot path. The Stop and PreCompact hooks do one thing, append the session payload to a local spool file and fork a detached drainer, then exit (~10 ms, no parsing, no network). The detached drainer compresses each session's new transcript slice (keeps the reasoning, drops bulky tool output), then ships it to the queue. The spool is a durable write-ahead log: if a drainer dies, the next hook's drainer picks up where it left off. Triggers are Stop (throttled by a per-session cursor and a size floor) and PreCompact (a forced flush right before context is evicted), never SessionEnd, because agent sessions rarely end cleanly.

Secrets are redacted on the way in. Sessions are full of pasted keys, and a capture is read by the archivist's model, so the server strips well-known key formats (cloud, payment, model-provider and source-host keys, private key blocks) and secret-looking KEY=value assignments before a capture is stored. Anything a pattern cannot recognize still goes through, so treat a pasted secret as exposed and rotate it.

Leveled compaction (LSM-style). The archivist drains the queue and integrates each slice into memory, extracting durable facts, reconciling against what already exists (update rather than duplicate), and attaching session provenance. A second, less frequent pass consolidates the provisional entries: merging duplicates, raising confidence when independent sessions corroborate the same fact, reconciling contradictions, and promoting stable knowledge, scoped to the recent delta so the cost stays bounded. Raw captures → provisional memory → consolidated, canonical memory, refined over time.

The net effect: memory quality is decoupled from write volume. Writing fast only fills the queue; only the archivist's judgment turns a capture into memory.