<!-- Full-page Markdown export (rendered HTML → GFM). Source: https://neotoma.io/zh/inspector/timeline Generated: 2026-04-27T12:48:35.203Z --> # Inspector, Timeline & interpretations The Timeline view turns Neotoma's per-row history into a single chronological stream. Anything that produced a write, a new observation, a correction, a new edge, a merge, an interpretation, shows up here, scoped to the operator's `user_id`. It's the fastest way to answer "what just happened?" without writing a query. ## Timeline stream Each event row carries a timestamp, an event kind (`store`, `correct`, `link`, `merge`, `interpret`), the affected entity, the agent that did it, and the trust tier. Events are clickable and deep-link to the underlying observation, relationship, or entity row. inspector.neotoma.io/timeline?range=24h Inspector Neotoma Dashboard Conversations Turns Compliance Activity Feedback Entities Observations Sources Relationships Graph Explorer Schemas Timeline Interpretations Agents Agent grants Settings Timeline Last 24h · 1,021 events store (612)correct (118)link (231)merge (4) 12:41storetransaction · Subscription · Vercelclaude-code 12:38correcttransaction · category → subscriptionoperator 12:30storeagent\_message · turn 4 (user)claude-code 12:30linkPART\_OF · turn 4 → conversation Q2 reviewclaude-code 12:21interpretinterpretation · spend pattern · Aprilingest-pipeline 11:08storereceipt · vercel-2026-04claude-code 11:08linkEMBEDS · receipt → file\_asset (pdf)claude-code 10:55mergecontact · Sarah Park ← Sarah P.operator Reverse-chronological event stream with quick filters by kind, entity type, agent, and trust tier. ◆ ## Range and ranges The range chip switches between rolling windows (last hour, 24h, 7d, 30d) and absolute ranges. The same query is reachable programmatically via `list_timeline_events` on the API. ◆ ## Filters - **Event kind**, store, correct, link, merge, interpret. - **Entity type**, same enum as the entity list filter. - **Agent**, by signed identity (AAuth thumbprint), `clientInfo.name`, or free-form `agent_label`. - **Trust tier**, hardware, software, unverified, anonymous. ◆ ## Interpretations Interpretations are derived entities, synthesised summaries (legal\_research, competitive\_analysis, market\_research, report, analysis-style notes) produced by an agent on top of raw observations. Inspector exposes them as a separate section (`/interpretations`) with their inputs, conclusions, caveats, and the source observations they cite. inspector.neotoma.io/interpretations Inspector Neotoma Dashboard Conversations Turns Compliance Activity Feedback Entities Observations Sources Relationships Graph Explorer Schemas Timeline Interpretations Agents Agent grants Settings Interpretations 48 derived analyses · last 30 days research\_date desc Q2 spend pattern (subscriptions)reportApr 24 cites 124 observations·claude-code Vercel vs Netlify cost comparisoncompetitive\_analysisApr 22 cites 38 observations·cursor-agent Sarah Park, relationship summaryanalysisApr 21 cites 17 observations·claude-code Email triage backlog (week 16)reportApr 20 cites 92 observations·ingest-pipeline Interpretations are synthesised entities; each links to the source observations and to the entity it interprets. ◆ ## Replay Selecting a range and pressing *Replay* walks the events in order, with the current snapshot visible alongside. This is useful when investigating "why did this entity look like X at 11:02 but Y at 11:08?", the timeline + reducer transparency together answer that without a database query. See [Observations & sources](/inspector/observations-and-sources) for the underlying immutable history that powers replay.