---
title: "Read Engagement Timeline"
method: GET
path: "/engagements/{engagement_id}/timeline"
tags: ["authenticated", "engagements"]
---

# Read Engagement Timeline

`GET /engagements/{engagement_id}/timeline`

## Request

- Base URL: `https://api-docs.upsmith.com` (the document declares no server; this is the origin it was published from)
- URL: `https://api-docs.upsmith.com/engagements/{engagement_id}/timeline`
- Auth: one of:
  - API key in header `x-api-key`
  - HTTP bearer

Paginated, chronological event stream for an engagement.

``cursor`` is the opaque ``{timestamp}|{event_id}`` token from the prior
page's ``cursor``; ``reverse=True`` loads the most-recent page first (used
on initial open) then pages backward into older events.

## Path parameters

- `engagement_id` string, required

## Query parameters

- `cursor` string, nullable
- `reverse` boolean
- `include_ai_response_changes` boolean

## Headers

- `organization-id` string, nullable
- `x-user-id` string, nullable
- `timezone` string
- `x-actor-email` string, nullable
- `x-request-source` string, nullable
- `x-operation-id` string, nullable
- `x-commit-message` string, nullable
- `x-confirmation-id` string, nullable
- `x-chat-id` string, nullable

## Response `200`

Successful Response

- PageUnionTimelineEventPublicAiResponseChangeTimelinePublic
  - `items` union[], required
    - union
      - TimelineEventPublic
        - `id` string, required
        - `organization_id` string, required
        - `engagement_id` string, required
        - `origin` 'internal' | 'external', required
        - `role` 'user' | 'agent' | 'contact' | 'provider', required
        - `type` 'action' | 'interaction' | 'llm_response', required
        - `user_id` string, nullable
        - `created_at` string, date-time, required
        - `updated_at` string, date-time, required
        - `deleted_at` string, date-time, nullable
        - `user_initials` string, nullable
        - `interaction` TimelineEventInteractionPublic
          - `id` string, required
          - `channel_id` string, required
          - `contact_method_id` string, nullable
          - `image_id` string, nullable
          - `external_id` string, nullable
          - `twilio_message_status` string, nullable
          - `twilio_error_code` string, nullable
          - `telnyx_message_status` string, nullable
          - `telnyx_error_code` string, nullable
          - `metadata_json` object, nullable
          - `status` 'pending' | 'delivered' | 'received' | 'failed', required
          - `content` string, required
          - `translated_content` string, nullable
          - `deliver_after` string, date-time, nullable
          - `image` string, nullable
          - `image_content_type` string, nullable
          - `contact_method` ContactMethodPublic
            - `id` string, required
            - `type` 'phone' | 'email' | 'fax' | 'social', required
            - `social_platform` 'meta_messenger' | 'google_lsa' | 'instagram' | 'yelp' | 'thumbtack'
            - `phone_number` string, nullable
            - `email` string, nullable
            - `person_name` string, nullable
            - `person_role` string, nullable
        - `action` TimelineEventActionPublic
          - `id` string, required
          - `tool_module_id` string, required
          - `success` boolean, nullable
          - `type` 'booking' | 'booking_request'
          - `invocation_source` 'llm' | 'precall': Who invoked the tool that produced this action. ``LLM`` — the agent's model chose to call the tool mid-conversation. ``PRECALL`` — the system/operator ran the tool up front (before-turn hooks / pre-agent preload) and injected the result into the prompt; the model did not choose it. ``NULL`` means **legacy / unknown** — the row predates reliable tagging (ENG-4234) and is NOT necessarily LLM-invoked. Precalled tools were saved as actions before #4213 (2026-04-04), then not saved at all until ENG-4234, so historical rows are a mix we deliberately don't reconstruct. Readers must treat NULL as its own bucket, never fold it into ``LLM``. Only rows written after ENG-4234 shipped carry a trustworthy value.
          - `input` string, required
          - `output` string, required
          - `tool_module` ToolModulePublic
            - `id` string, required
            - `created_at` string, date-time, nullable
            - `updated_at` string, date-time, nullable
            - `deleted_at` string, date-time, nullable
            - `title` string, required
            - `description` string, required
            - `function` string, required
            - `disabled_reason` string, nullable
            - `read_only` boolean, required
        - `llm_response` EventLLMResponsePublic
          - `id` string, required
          - `phase` 'ModelRequest' | 'ModelResponse', required: Phase for persisted LLM transcript rows. Like other engagement ``StrEnum`` + ``sa.Enum(native_enum=False)`` columns, PostgreSQL stores the member **name** (``REQUEST`` / ``RESPONSE``) for ``llmphase``. The **value** strings are pydantic-ai message kinds (``ModelRequest`` / ``ModelResponse``) and are not written to this column.
          - `model` string, required
          - `content` union
            - object
            - unknown[]
              - …
            - string
          - `duration_ms` integer, nullable
      - AiResponseChangeTimelinePublic
        - `kind` 'ai_response_change'
        - `id` string, required
        - `engagement_id` string, required
        - `occurred_at` string, date-time, required
        - `is_ai_allowed_to_respond` boolean, required
        - `source` string, required
        - `reason` string, nullable
        - `operator_name` string, nullable
  - `total` integer, nullable
  - `cursor` string, nullable

## Other responses

- `422`: Validation Error

---

[API](https://skmtc.dev/upsmith/apis/upsmith-api.md) - [All operations](https://skmtc.dev/upsmith/apis/upsmith-api/llms.txt) - [OpenAPI document](https://skmtc.dev/upsmith/apis/upsmith-api/revisions/cce8b07a5787?raw)