@irisrun/inspect

See what your agent decided — read-only, snapshot-safe. A recorded session is a snapshot plus a journal; this reads it back as a decision/effect/marker timeline keyed by the stable sessionId. Nothing it derives re-enters replayed state, so inspecting a session can never change it — and re-inspecting the same journal bytes is byte-identical.

What it is

inspectSession(store, sessionId) reads the latest snapshot + journal tail from a StateStore, decodes each record, and returns a SessionInspection — governingDigest, snapshotUpTo, the InspectedRecord[] timeline, per-kind counts, and a terminal of finished | parked | open. The governing digest is re-derived LOCALLY from the post-snapshot tail (snapshot-safely), so the package depends on @irisrun/core ONLY. renderTimeline(inspection) formats it as one deterministic line per record.

This is the library API — not a CLI. (iris inspect is an image-only command.)

Use it

import { inspectSession, renderTimeline } from "@irisrun/inspect";

const inspection = await inspectSession(store, "s1");
console.log(renderTimeline(inspection));    // one line per decision/effect/marker

See docs/Inspecting a session.

Part of Iris — own, portable, verifiable state.