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

# Changelog

## July 22, 2026

### Integer Custom Enrichments

The `GET /customEnrichments` endpoint now returns enrichments with `type: "integer"` alongside the existing `categorical` type. Integer enrichments store numeric values and have `constraints.enum` always set to `[]`. They cannot be used in filters or aggregations.

## June 16, 2026

### Custom Enrichments API

You can now attach your own custom enrichments to mentions in PeakMetrics with two new endpoints:

* `GET /customEnrichments` — list the custom enrichments provisioned for your team, including each one's `enrichment_id`, `name`, `type`, and `constraints`.
* `PUT /mentions/customEnrichments` — push values for up to 100 mentions per request. Supports both standard and custom channel mentions, including historical mentions already in the platform.

Custom enrichments let you push external context about mentions into PeakMetrics, used alongside all other mention metadata in narrative analysis, summaries, and notifications, and available through the customer API and MCP server.

Contact [support@peakmetrics.com](mailto:support@peakmetrics.com) to set up custom enrichments!

## May 12, 2026

### Tweet Engagement Tracking Data on Mentions

The `GET /workspaces/{workspaceId}/mentions` endpoint now returns a new
`enrichments.engagement` object.

* `enrichments.engagement.history` is an array of timestamped snapshots (`measured_at`,
  `views`, `retweets`, `quotes`, `replies`, `favorites`, `bookmarks`) captured on a backoff
  schedule during the first week after a Tweet's published date, for Twitter / X mentions
  that have been opted into engagement tracking.
* For mentions that aren't being tracked — including mentions on channels that can't be
  tracked, like non-Twitter mentions — `history` is returned as an empty array.

See [Tracking Tweet Engagement](/help/enrichments-and-integrations/tracking-tweet-engagement) for details on how tracking is enabled and the refresh schedule.

## May 8, 2026

### Randomized Sampling on Mentions Endpoint

The `GET /workspaces/{workspaceId}/mentions` endpoint now supports an optional `randomization_key` query parameter. When provided, the endpoint returns a randomized sample (up to 500 mentions across all pages).

Key features:

* **Stable pagination** via `randomization_key` — reuse the same key across requests to paginate through a consistent randomized pool (cached for 5 minutes)
* All standard workspace mention filters are supported, and mentions are randomly pulled from the full time range

## May 1, 2026

### Sentiment Fields Deprecation Now in Effect

The deprecation of sentiment fields announced on March 23 is now live. The following fields are deprecated and return `0` for all responses:

* `/workspaces/{id}/mentions` — `enrichments.sentiment.polarity` now returns `0`
* `/workspaces/{id}/mentions` — `enrichments.sentiment.version` now returns `0`
* `/workspaces/{id}/narratives` — `aggregations.avgSentiment` now returns `0`

These fields remain in the response schema for backward compatibility but no longer contain meaningful data.

Additionally for `/workspaces/{id}/narratives`, `avgSentiment` is no longer accepted as an option for the `sort` query param.

## March 23, 2026

### Upcoming Deprecation of Sentiment Fields

<Info>
  For an upgraded sentiment metric, you can set up a **custom favorability smart category** tailored to your workspace. Contact [support@peakmetrics.com](mailto:support@peakmetrics.com) to get started.
</Info>

The generic sentiment model and its associated fields will be deprecated on **May 1, 2026**. After this date, the following changes take effect:

* The `/workspaces/{id}/mentions` endpoint will return `0` for `sentiment.polarity` and `sentiment.version`
* The `/workspaces/{id}/narratives` endpoint will return `0` for `aggregations.avgSentiment`

These fields will remain in the response schema for backward compatibility but will no longer contain meaningful data.