Interface: AgentArtifactsOptions

Defined in: src/core/agent/types.ts:69

The object form of AgentOptions.artifacts (9.22.0): the store plus its operator dials. placement cannot be spelled without store — a threshold with nowhere to put what it catches would be configuration that lies, and the shape refuses it before a runtime check has to.

Properties

placement?

readonly optional placement?: ArtifactPlacement

Defined in: src/core/agent/types.ts:81

The placement threshold. A tool result whose finalized text exceeds maxInlineChars is checked into the store (kind tool-result/<toolName>) and the model reads the claim ticket instead. Judged AFTER the tool's own resultCeiling (the author's refusal comes first) and BEFORE the agent-level maxToolResultChars truncation net (which then measures the ticket, so it should rarely fire). Omitted → results are never measured and never placed, exactly as before.

---

recordings?

readonly optional recordings?: boolean | AgentRecordingsOptions

Defined in: src/core/agent/types.ts:119

Check each completed run's RECORDING into the store (9.26.0), so a screen can replay the turn later without the deployment inventing a place to keep it.

true for the default naming, or { label } to name them yourself. Off by default: unset, no recorder is attached, no events are captured, and the run is byte-identical to every earlier release.

── What lands, and where ─────────────────────────────────────────────── The recordRun contract exactly — { snapshot, events, structure }, the three things a viewer needs and the shape observeRecording() consumes — minted as kind 'recording/run' in the RUN's own artifact scope, with origin.runId joining it back to the trace. The existing wire ops serve it: { op: 'artifact-get', ref } returns the recording, and no new operation was needed for any of it. Retention rides the store, so recordings age out under the same ttl and byte budget everything else does.

── The cost, stated rather than discovered ───────────────────────────── Recording a run means an event tail and a boundary recorder for its duration, and the mint is one store write on the way OUT of run() — the answer is fully composed before the write begins and the write can never change it, but run() does return after it rather than before. That is deliberate: a fire-and-forget write is a recording lost whenever the container exits with the reply, which is exactly the deployment that wants this most.

A mint that FAILS degrades to the old behaviour — the answer is returned unchanged, and the failure lands on the record as agentfootprint.artifacts.refused. A run never fails because its recording could not be filed.

Nothing is minted for a run that paused (there is no completed run yet) or threw. A resumed run mints when it completes, and its recording covers the RESUMED run — which is what the recorder saw.

---

store

readonly store: ArtifactStore

Defined in: src/core/agent/types.ts:71

The claim-check store — same seam as the bare ArtifactStore form.