Activity Feeds V3 is in closed alpha — do not use it in production (just yet).

Intro & Defaults

Creating Feed Groups & Defaults

There are several feed groups setup by default.

Group	Description
`user`	A feed setup for the content a user creates. Typically you add activities here when someone writes a post
`timeline`	The timeline feed is used when you’re following. So if user Charlie is following John, timeline:charlie would follow user:john
`foryou`	A version of the timeline feed that adds popular content, and priorities popularity over recency
`notification`	A notification feed. Think of the bell icon you see in most apps

You can also create your own feed group. Here’s how to create a feed group using the API:

// Note: Feed group management is typically done server-side
// This is an example of the configuration structure
let feedGroupConfig: [String: Any] = [
    "id": "myid",
    "activity_processors": [["type": "topic"]], // activity processors do post processing on activities. AI topic analysis, etc.
    "activity_selectors": [["type": "following"]], // control how activities end up in this feed. direct, following, search, querying etc.
    "ranking": ["type": "time"], // how activities are ranked
    "aggregation": ["type": "count"], // group activities together
    "notification": ["type": "count"], // keep track of seen and read state on notification feeds
    "custom": ["description": "My custom feed group"] // anything else to store on the feed group
]

const view = await serverClient.feeds.createFeedView({
  view_id: uuidv4(),
  activity_selectors: [
    {
      type: "following",
    },
  ],
  ranking: { type: "time" },
  // Group by activity type
  aggregation: { type: "count" },
});

const response = await serverClient.feeds.createFeedGroup({
  feed_group_id: "myid",
  default_view_id: view.feed_view.view_id,
  activity_processors: [
    {
      type: "topic",
    },
  ],
  notification: {
    track_read: true,
    track_seen: true,
  },
  custom: {
    description: "My custom feed group",
  },
});

For You Feeds

Many apps want to have a “for you” or personalized feed. There are a couple benefits to a personalized feed:

Works well even if your users don’t spend much time setting up follows
Can be a mechanism to discover new content or things to follow

Stream offers a few built-in methods to create a for you feed and gives you the API access to do more advanced customization if needed.

Activity Ranking

When ranking activities you can specify a ranking formula. This can combine an age decay with popularity and other factors.

// Note: This is typically configured server-side
let rankingConfig = [
    "type": "time",
    "score": "decay_linear(time) * popularity"
]

let feedGroupConfig: [String: Any] = [
    "id": "timeline",
    "ranking": rankingConfig
]

const view = await serverClient.feeds.createFeedView({
  view_id: uuidv4(),
  ranking: { type: "time", score: "decay_linear(time) * popularity" },
  activity_selectors: [
    {
      type: "following",
    },
  ],
});

const response = await serverClient.feeds.createFeedGroup({
  feed_group_id: "mytimeline",
  default_view_id: view.feed_view.view_id,
});

Interest Based “For You” Style Feed

This next example is a bit more complicated. It uses an activity processor to add topic data to activities and ranks content you’re likely to engage with higher in the feed.

// Note: This is typically configured server-side

// Run the activity processors to analyse topics for text & images
let imageProcessor = ["type": "image_topic"]
let textProcessor = ["type": "text_topic"]
let userFeedConfig: [String: Any] = [
    "id": "user",
    "activity_processors": [imageProcessor, textProcessor]
]

// Activity selectors change which activities are included in the feed
// The default "following" selectors gets activities from the feeds you follow
// The "popular" activity selectors includes the popular activities
// And "interest" activities similar to activities you've engaged with in the past
// You can use multiple selectors in 1 feed
let activitySelectors = [
    ["type": "popular"],
    ["type": "following"],
    ["type": "interest"]
]

// Rank for a user based on interest score
// This calculates a score 0-1.0 of how well the activity matches the user's prior interest
let rankingConfig = [
    "type": "interest",
    "score": "decay_linear(time) * interest * decay_linear(popularity)"
]

let timelineConfig: [String: Any] = [
    "id": "timeline",
    "ranking": rankingConfig,
    "activity_selectors": activitySelectors
]

// Read the feed
// Activities will include following, popular and similar (via interest) activities
// Sorted by time decay, interest and popularity
let forYouFeed = client.feed(group: "timeline", id: "thierry")
let response = try await forYouFeed.getOrCreate()

const view = await serverClient.feeds.createFeedView({
    view_id: uuidv4(),
    ranking: { type: 'expression', score: 'decay_linear(time) * popularity' },
    // Activity selectors change which activities are included in the feed
    // The default "following" selectors gets activities from the feeds you follow
    // The "popular" activity selectors includes the popular activities
    // And "interest" activities similar to activities you've engaged with in the past
    // You can use multiple selectors in 1 feed
    activity_selectors: [
        {
            type: 'popular',
        },
        {
            type: 'following',
        },
        {
            type: 'interest',
        },
    ],
    // Rank for a user based on interest score
    // This calculates a score 0-1.0 of how well the activity matches the user's prior interest
    ranking: {
        type: "interest",
        score: "decay_linear(time) * interest * decay_linear(popularity)"
    },

});

const response = await serverClient.feeds.createFeedGroup({
    feed_group_id: 'mytimeline',
    default_view_id: view.feed_view.view_id,
    // Run the activity processors to analyse topics for text & images
    activity_processors: [
        {type: 'image_topic'},
        {type: 'text_topic'}
    ]
});

const forYouFeed = client.feeds.feed(group: "mytimeline", id: "thierry")
const response = await forYouFeed.getOrCreate({ user_id: 'thierry' })

Additional Ranking

There are a few ways that you can extend the ranking in feeds. This allows you pass in any insights you have about the user.

// Pass external ranking data when reading the feed
let query = FeedQuery(
    group: "timeline",
    id: "thierry",
    externalRanking: [
        "user_engagement_score": 0.8,
        "content_preference": 0.9,
        "time_of_day_bonus": 1.2
    ]
)
let feed = client.feed(for: query)
let feedData = try await feed.getOrCreate()

// Pass external ranking data when reading the feed
const externalRanking = {
  user_engagement_score: 0.8,
  content_preference: 0.9,
  time_of_day_bonus: 1.2,
};

const response = await feed.getOrCreate({
  external_ranking: externalRanking,
});

// Pass external ranking data when reading the feed
const externalRanking = {
  user_engagement_score: 0.8,
  content_preference: 0.9,
  time_of_day_bonus: 1.2,
};

const response = await feed.getOrCreate({
  external_ranking: externalRanking,
  user_id: "<user id>",
});

You can also rank on custom data in the activity:

let activity = try await feed.addActivity(
    request: .init(
        custom: [
            "quality_score": 0.95,
            "engagement_prediction": 0.8
        ],
        text: "Great content",
        type: "post"
    )
)

const activity = await feed.addActivity({
  type: "post",
  text: "Great content",
  custom: {
    quality_score: 0.95,
    engagement_prediction: 0.8,
  },
});

const activity = await client.feeds.addActivity({
  fids: [feed.fid],
  type: "post",
  text: "Great content",
  custom: {
    quality_score: 0.95,
    engagement_prediction: 0.8,
  },
  user_id: "<user id>",
});

Ranking

The ranking system allows you to customize how activities are ordered in your feeds. Here are some common ranking strategies:

Time-based Ranking

// Simple time-based ranking (newest first)
let timeRanking = [
    "type": "time",
    "direction": "desc"
]

// Simple time-based ranking (newest first)
const timeRanking = {
  type: "time",
  direction: "desc",
};

Popularity-based Ranking

// Rank by popularity (likes, comments, shares)
let popularityRanking = [
    "type": "popularity",
    "score": "likes + comments * 2 + shares * 3"
]

// Simple time-based ranking (newest first)
const popularityRanking = {
  type: "popularity",
  score: "likes + comments * 2 + shares * 3",
};

Hybrid Ranking

// Combine time decay with popularity
let hybridRanking = [
    "type": "hybrid",
    "score": "decay_linear(time, 24h) * (likes + comments + shares)"
]

// Combine time decay with popularity
const hybridRanking = {
  type: "hybrid",
  score: "decay_linear(time, 24h) * (likes + comments + shares)",
};

Interest-based Ranking

// Rank based on user interests and engagement history
let interestRanking = [
    "type": "interest",
    "score": "decay_linear(time) * interest_match * engagement_prediction"
]

// Rank based on user interests and engagement history
const interestRanking = {
  type: "interest",
  score: "decay_linear(time) * interest_match * engagement_prediction",
};

Aggregation & Notification Feeds

Aggregation groups similar activities together, which is useful for notification feeds and reducing noise.

Aggregation Types

// Count aggregation - shows "John and 5 others liked your post"
let countAggregation: [String: Any] = [
    "type": "count",
    "group_by": ["activity_id", "reaction_type"]
]
// Time-based aggregation - groups activities within a time window
let timeAggregation = [
    "type": "time",
    "window": "1h"
]
// User-based aggregation - groups activities by user
let userAggregation: [String: Any] = [
    "type": "user",
    "group_by": ["user_id"]
]

// Count aggregation - shows "John and 5 others liked your post"
const countAggregation = {
    type: "count",
    group_by: ["activity_id", "reaction_type"]
}

// Time-based aggregation - groups activities within a time window
const timeAggregation = {
    type: "time",
    window: "1h"
}

// User-based aggregation - groups activities by user
const userAggregation = [
    type: "user",
    group_by: ["user_id"]
]

Notification Feed Example

let notificationFeed = client.feed(group: "notification", id: "john")
let notifications = try await notificationFeed.getOrCreate()

// Built in notification group
const notificationFeed = client.feed(group: "notification", id: "john")

// Read notifications
const notifications = (await notificationFeed.getOrCreate({
    limit: 20,
})).aggregated_activities;

// Built in notification group
const notificationFeed = client.feed(group: "notification", id: "john")

// Or create your own config
const view = await serverClient.feeds.createFeedView({
    view_id: uuidv4(),
    ranking: { type: 'time', direction: 'desc' },
    aggregation: {
        type: "count",
        group_by: ["activity_id", "reaction_type"]
    }
});
const response = await serverClient.feeds.createFeedGroup({
    feed_group_id: 'mynotification',
    default_view_id: view.feed_view.view_id,
    notification: {
        track_read: true,
        track_seen: true,
    }
});

// Read notifications
const notifications = (await notificationFeed.getOrCreate({
    limit: 20,
    user_id: '<user_id>'
})).aggregated_activities;

Marking Notifications as Read

// Mark specific notifications as read
try await notificationFeed.markActivity(
    request: .init(
        markRead: ["notification_1", "notification_2"]
    )
)
// Mark all notifications as read
try await notificationFeed.markActivity(
    request: .init(
        markAllRead: true
    )
)

// Mark specific notifications as read
await notificationFeed.markActivity({
  mark_read: ["notification_1", "notification_2"],
});

// Mark all notifications as read
await notificationFeed.markActivity({
  mark_all_read: true,
});

// Mark specific notifications as read
await notificationFeed.markActivity({
  mark_read: ["notification_1", "notification_2"],
  user_id: "<user id>",
});

// Mark all notifications as read
await notificationFeed.markActivity({
  mark_all_read: true,
  user_id: "<user id>",
});

Follows

Aggregation