Observability

The observability screens (the Data section of the sidebar) answer three questions in order: did my data arrive, what did it change, and what did agents end up reading? Use them to verify ingestion, trace state changes, and audit agent-facing output without inferring behavior from derived state alone.

​

Logs

The log browser shows raw logs exactly as ingested. A volume histogram over your selected time range makes gaps and spikes visible at a glance. Narrow the stream with filters:

• Event type category: symptom reports, labs, medication, and so on.
• Processing state: where the log is in the pipeline.
• Patient and source: a single person or a single integration.
• Search: free text, including individual log IDs.

Every log row has a processing state:

• Completed: ingestion and downstream processing finished.
• Running: pipeline still processing this log.
• Unprocessed: persisted but not yet run through state workflows (common during batch backfills).
• Failed: processing errored; open the row for details.

Open any row to inspect the full payload as JSON and trace the individual submission. You can also select multiple rows to delete stale or unprocessed logs, useful when cleaning up after a misconfigured integration or a bad backfill.

​

Events

Events is a timeline of state-affecting activity: each entry is an event derived from one or more logs, with the state changes it caused. Where Logs shows the raw request body, Events shows the effect. For each event you can:

• View the state diff: before and after values for every module the event touched.
• Filter by event type, patient, time range, or how many state updates an event produced.
• Jump to the source logs that produced the event.

Use it to step forward and back through what changed in a patient’s state and to answer “why does this module say that?”

​

Views

Inspect generated patient views over time: every snapshot, which template produced it, and when. Each snapshot opens into three tabs:

• Blocks: per-block results with trust and confidence badges.
• Evolution: how the view’s content changed across builds.
• Provenance: links from each block back to the state modules and logs that fed it.

Reach for this screen when a block is empty, stale, or surprising relative to the underlying modules; the provenance tab usually pinpoints the cause.

​

Population views

Where patient views cover one person, population views aggregate a cohort: org-level template results with aggregate tiles and per-step inspection, filterable by time range. Use them to audit cohort-level summaries the same way Views audits a single patient.