Deploy Feed Configs

A feed config turns saved algorithms into a served feed. Use it after you have built and validated algorithms in the Algo Builder.

Console Steps

1. Open Feed in Console.
2. Create a new feed config or select an existing one.
3. Add one or more saved algorithms to Main algorithms.
4. Set weights when using more than one main algorithm.
5. Select primary and secondary fallback algorithms if needed.
6. Set cache age values.
7. Configure exclusion rules.
8. Save the config.
9. Use the test panel to verify served output.

Main Algorithms

Main algorithms are the normal path for the feed. Start with one main algorithm. Add multiple algorithms only when each one contributes a distinct candidate pool.

Example mix:

{
  "main_algos": [
    { "algo_id": "<primary-algo-id>", "weight": 0.7 },
    { "algo_id": "<secondary-algo-id>", "weight": 0.3 }
  ]
}

Weights should add up to the mix you want. A 0.7 / 0.3 mix means roughly 70% of candidates should come from the first algorithm and 30% from the second, subject to available results.

Fallbacks

Fallbacks prevent empty feeds.

Fallback algorithms should be less restrictive than the main algorithm. If the main algorithm filters by a narrow account, label, or market, the fallback should loosen that condition.

Cache Settings

Console exposes cache controls so the feed can balance freshness and cost.

Use shorter cache windows for fast-moving markets or notifications. Use longer windows for stable discovery feeds.

Exclusion Rules

Exclusion rules keep the served feed from repeating items a user has already seen or should not see again.

Common rules:

• Hide items shown in the last N seconds.
• Hide records with a matching field from a prior result.
• Hide consumed, dismissed, or already-delivered items when those fields exist.

Use the test panel after every exclusion change. A too-broad exclusion can make the feed appear empty even when the underlying algorithms return hits.

Serve And Test

The algorithm preview answers: does this algorithm return candidates?

The feed config test answers: does production serving return the right feed after weights, fallbacks, cache, and exclusions?

Run both before shipping.

Example: Production Feed Config

{
  "feed_type": "minifeed",
  "main_algos": [
    { "algo_id": "<primary-algo-id>", "weight": 0.7 },
    { "algo_id": "<secondary-algo-id>", "weight": 0.3 }
  ],
  "primary_fallback": 180,
  "secondary_fallback": 181,
  "main_max_cache_age_s": 7200,
  "primary_max_cache_age_s": 7200,
  "secondary_max_cache_age_s": 600,
  "exclusion_items_shown_last_n_seconds": 3600
}

Replace the algorithm IDs with the saved IDs from your Console account.

Pattern: Mixing Feed Algorithms

Use feed configs to combine saved algorithms by role, not by ID. A strong production setup usually has:

• One broad main algorithm.
• One or two specific main algorithms.
• A broader primary fallback.
• A very broad secondary fallback.

{
  "feed_type": "minifeed",
  "main_algos": [
    { "algo_id": "<recent-trades-algo-id>", "weight": 0.5 },
    { "algo_id": "<whale-trades-algo-id>", "weight": 0.25 },
    { "algo_id": "<watchlist-algo-id>", "weight": 0.25 }
  ],
  "primary_fallback": "<broad-market-algo-id>",
  "secondary_fallback": "<cold-start-algo-id>",
  "main_max_cache_age_s": 7200,
  "primary_max_cache_age_s": 7200,
  "secondary_max_cache_age_s": 600
}

Use a broad market or fresh-activity algo as fallback. Avoid using a highly personalized watchlist algo as the final fallback because it can be empty for cold-start users.

Next

• Build algorithms in Building Algorithms.
• Learn feed concepts in Feed Management.
• Check the served feed with Deploy Serving.