# Data Model & Step Types Reference

An `Interaction` response contains `steps`, an array of typed step objects
representing a structured timeline of the interaction turn. Read the current
step `type` rather than assuming the last step is text — the trailing step may
be a `function_call` or a `thought`.

## Step Types

**User steps:**

*   `user_input`: User input (text, audio, multimodal). Contains a `content`
    array. (This is why REST input payloads use `"type": "user_input"`, **not**
    `"role": "user"`.)

**Model/server steps:**

*   `model_output`: Final model generation. Contains a `content` array with
    `text`, `image`, `audio`, etc. (REST responses use `"type": "model_output"`,
    **not** `"role": "model"`.)
*   `thought`: Model reasoning / chain of thought. Has a `signature` field and
    optional `summary`.
*   `function_call`: Tool call request, with flat `id`, `name`, and `arguments`
    fields (there is **no** nested `tool_calls` list).
*   `function_result`: Tool result you send back, with `call_id`, `name`, and
    `result` fields.
*   `google_search_call` / `google_search_result`, `code_execution_call` /
    `code_execution_result`, `url_context_call` / `url_context_result`,
    `mcp_server_tool_call` / `mcp_server_tool_result`, `file_search_call` /
    `file_search_result`: built-in and remote tool steps.

## Content types (inside the `content` array on `model_output` and `user_input` steps)

*   `text`: Text content (`text` field).
*   `image` / `audio` / `document` / `video`: Content with `data`, `mime_type`,
    or `uri`.

## Convenience accessors

*   `output_text`: The combined text from the trailing `model_output` steps.
    Prefer this over hand-walking `steps[-1].content[0].text`, which breaks when
    the last step is a tool call or a thought.
*   `output_image` / `output_audio`: The last image or audio generated by the
    model in the response (`data` base64 + `mime_type`).

## Streaming Event Types

| Event                       | Description                                                                                                       |
| --------------------------- | ----------------------------------------------------------------------------------------------------------------- |
| `interaction.created`       | Interaction created; includes metadata.                                                                           |
| `interaction.status_update` | Interaction-level status change.                                                                                  |
| `step.start`                | A new step begins. Contains the step `type` and initial metadata.                                                 |
| `step.delta`                | Incremental data for the current step. Contains a typed `delta` object (e.g. `delta.type == "text"` with `delta.text`). |
| `step.stop`                 | The step is complete. Contains `index`.                                                                           |
| `interaction.completed`     | Interaction finished. Contains final `usage`.                                                                     |

## Delta Types

| Delta Type          | Parent Step    | Description                                  |
| ------------------- | -------------- | -------------------------------------------- |
| `text`              | `model_output` | Incremental text token.                      |
| `audio`             | `model_output` | Audio chunk (base64).                        |
| `image`             | `model_output` | Image chunk (base64).                        |
| `thought_summary`   | `thought`      | Thinking summary text.                       |
| `thought_signature` | `thought`      | Opaque signature for thought verification.   |

**Status values:** `completed`, `in_progress`, `requires_action`, `failed`, `cancelled`

## Storage & retention

Interactions are stored by default (`store=True`, retained for 7 days), which
enables stateful features like `previous_interaction_id` and background
execution. Passing `store=False` disables server-side retention (Zero Data
Retention compliant) and therefore also disables `previous_interaction_id` and
`background` — in that mode you must pass the full conversation history in
`input` on each turn.