> ## Documentation Index
> Fetch the complete documentation index at: https://docs.verial.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Interactions

> Recorded evidence captured on each sandbox during a rollout.

An **interaction** is the evidence a [Sandbox](/guides/concepts/sandboxes) records while the agent drives a rollout. Every request to a FHIR store, every HL7 outbound, every portal form submit, every voice turn, every uploaded file, every X12 response: Verial writes it down. The [verification engine](/guides/concepts/verification) reads these interactions when it runs each [Criterion](/guides/concepts/criteria) after the task run completes.

## Evidence by Simulator

Each simulator type produces its own evidence shape. The verification engine dispatches to a check implementation keyed by `assertion.assert`, and each check pulls evidence from the matching source.

| Simulator        | Evidence                                                                                                                                          |
| ---------------- | ------------------------------------------------------------------------------------------------------------------------------------------------- |
| **FHIR**         | HTTP request/response log: method, path, request body, status code, response body. Verification runs FHIR searches against the final store state. |
| **HL7**          | Outbound HL7v2 messages recorded as `hl7_outbound` sandbox events, with the full message payload (MSH, PID, PV1, OBX, etc.).                      |
| **Voice**        | Call turns with speaker (`agent` / `caller`) and transcribed text, plus the full recording.                                                       |
| **Fax**          | Inbound or outbound fax document, with OCR text extracted for assertion.                                                                          |
| **Portal**       | Sandbox events per action: form submits, patient searches, auth submissions, with the submitted payload and the resulting state row.              |
| **Files / SFTP** | Uploaded file metadata (path, size) plus the raw content in object storage.                                                                       |
| **X12**          | Submitted and response records per transaction (270/271/276/277/278).                                                                             |
| **CDS Hooks**    | Hook invocations and the cards returned by the agent.                                                                                             |
| **Message**      | Outbound SMS/text messages with the rendered body.                                                                                                |

## How the Verification Engine Reads Interactions

```mermaid theme={null}
graph LR
  R["Rollout completes"]:::input --> I["Interactions in sandbox state"]:::state
  I --> D["Dispatch on assertion.assert"]:::engine
  D --> C["Check reads matching evidence"]:::engine
  C --> O["Criterion Run<br/>(passed, score, evidence)"]:::output

  classDef input fill:#e0f2fe,color:#1e3a5f,stroke:#3b82f6
  classDef state fill:#fef3c7,color:#78350f,stroke:#f59e0b
  classDef engine fill:#2563eb,color:#fff,stroke:#1d4ed8
  classDef output fill:#f0fdf4,color:#166534,stroke:#22c55e
```

For each criterion on the task, the engine:

1. Reads `assertion.assert` to pick a check implementation.
2. Pulls the relevant evidence from the sandbox (a FHIR search against the store, HL7 outbound rows, portal state rows, voice turns, SFTP objects, X12 responses).
3. Runs the typed assertion against that evidence.
4. Writes a [Criterion Run](/api-reference/resources/criterion-runs) with `passed`, `score`, `details`, and the evidence it considered.

See [Verification](/guides/concepts/verification) for the full dispatch table and scoring rules.

## Reading Interactions

Interactions surface in two places:

* **Per sandbox**: `GET /sandboxes/{id}/events` returns the raw event log for a sandbox. Useful for debugging a rollout or authoring new criteria from real traces.
* **Per criterion run**: `GET /criterion-runs/{id}` returns the specific evidence the check considered for that criterion, with field-level diffs where applicable.

During a scored benchmark run, per-field evidence is omitted from completion responses so the agent cannot learn the rubric. Fetch the full evidence later from `GET /criterion-runs/{id}`.

## Retention

Interactions persist after a playground is torn down. Teardown releases the live resources (phone numbers, FHIR stores, portal users) but keeps every recorded event so you can review evidence, debug failed criteria, and compare rollouts across benchmark runs.

## Next Steps

<CardGroup cols={2}>
  <Card title="Verification" icon="scale-balanced" href="/guides/concepts/verification">
    How interactions feed the scoring engine.
  </Card>

  <Card title="Sandboxes API" icon="code" href="/api-reference/resources/sandboxes">
    Read the raw event log for any sandbox.
  </Card>
</CardGroup>