> ## Documentation Index > Fetch the complete documentation index at: https://docs.neuronsearchlab.com/llms.txt > Use this file to discover all available pages before exploring further. # Pipeline > The stages a recommendation request runs through, and how to tune each one for a context. Every recommendation request runs through a **pipeline**: an ordered set of stages that retrieve candidates, score them, apply your rules and shape the final list. Each context has its own pipeline, created with its recipe, so a homepage feed and a "similar items" rail can work differently. Open a pipeline from a context's recipe, or from **Build a pipeline** on the Dashboard. The editor shows each stage as a card that you can turn on or off and configure. When opened from a context, the editor is locked to that context's pipeline. ## Stages Requests pass through the stages in this order. Turning a stage off skips it entirely. | Stage | Shown as | What it does | | - | - | - | | **Candidate generation** | Candidate Generation | Retrieves candidates from the context's sources, mainly the nearest items in the model's embedding index | | **Feature enrichment** | Signals | Attaches user, item and context signals used by later stages | | **Scoring** | Scoring | Scores candidates and applies discovery settings | | **Learned ranking** | Learned Reranker | A second-stage model that reorders candidates using behavioural features | | **Catalogue intelligence** | Catalog Signals | Blends item-to-item relationships and metadata similarity into scores | | **Intelligence layer** | LLM Reranker | Reranks the top of the list with a language model or cross-encoder against several objectives | | **Rules** | Rules Engine | Applies your [rules](https://docs.neuronsearchlab.com/console/rules.md) | | **Ranking** | Final Sort | Combines scores into the final order | | **Post-processing** | Output Shaping | Removes duplicates, spreads out categories and trims to the page size | ## Stage settings ### Candidate generation * **Candidate limit**: how many items to retrieve before scoring. Higher values widen the candidate pool and can increase latency. Evaluate a larger pool if results lack variety. * **Query / search retrieval**: for search requests, adds keyword matching alongside meaning-based matching. Choose **Reciprocal rank fusion** (rank-based and stable) or **Weighted normalised scores**, set the **semantic** and **keyword** weights, and pick the **keyword fields** (built-in fields and metadata keys) to match against. ### Scoring | Setting | Effect | | - | - | | **Freshness boost** | Score boost by item age: full for items under 7 days, half for 7 to 30 days | | **Impression fatigue** | Demotes items served repeatedly without meaningful engagement, judged by your event weights. Set **Strength** and the **Serve cap** | | **Trending backfill** | When personalised scores are weak, adds globally trending items so the list stays useful | | **Exploration slots** | Reserves a percentage of slots for items outside the user's usual profile | | **Search intent** | Leans recommendations towards what the user searched for recently. See [Searches are events](https://docs.neuronsearchlab.com/integrate/events.md#searches-are-events) | ### Learned ranking Choose the ranker type: | Ranker | Use it when | | - | - | | **XGBoost** | The default: a fast gradient-boosted ranker over behavioural, freshness and affinity features | | **MMoE** | Several objectives share one model (multi-gate mixture of experts) | | **PLE** | Objectives need stronger separation (progressive layered extraction) | **Ranking facets** tells the ranker which metadata fields to compute topic affinity over, such as `category` or `brand`. The first non-empty field on each item is used. Leave it empty to use standard category and genre keys. ### Catalogue intelligence Blends relationships between items (similar, complementary, substitute, same collection, frequently used together, next best and others) into model scores. * **Catalogue influence**: how far these signals can shift scores, capped at 50% so model relevance stays primary. * **Relationship types** and **sources**: which relationships may be used. * **Max relationship candidates**: how many relationship-derived candidates may enter per request. ### Intelligence layer An optional reranker for the top of the list. | Setting | Effect | | - | - | | **Model** | A managed cross-encoder (lowest latency), a self-hosted cross-encoder, or a language-model listwise reranker (slower, reasons over several objectives) | | **Objective** | **Balanced** and **Personalised** lean on relevance, **Discovery** favours diversity and freshness, **Compliance** prioritises policy. **Custom** shows a slider per objective | | **Depth** | How many top candidates are reranked. A greater depth considers more candidates and can increase latency; compare quality before raising it | | **Latency budget** | A hard timeout. If the reranker does not answer in time, the upstream order is kept | | **Fallback on error** | On by default: keep the upstream order if the reranker fails. Off makes the request fail instead | | **Explanations** | Off, brief or verbose reasons per item, shown in [Explainability](https://docs.neuronsearchlab.com/console/explainability.md) | ### Rules Turn this stage off to serve pure model ranking with no business rules, for example as the control arm of an [experiment](https://docs.neuronsearchlab.com/console/experiments.md). ### Post-processing * **Diversity**: limits consecutive items from the same category, pushing extras further down. * **Deduplicate**: removes duplicate items, keeping the highest-scored one. * **Page size**: how many items the response returns. Every earlier stage works on the full candidate set; the trim happens here. ## Video understanding If [Twelve Labs video enrichment](https://docs.neuronsearchlab.com/console/integrations.md#video-enrichment) is enabled for your workspace, the pipeline can use the stored video analysis for candidate vectors, with **Fallback to base embeddings** keeping recall stable for videos not yet enriched. ## Defaults A new context's pipeline comes from its template. A workspace with no pipeline at all uses defaults: all stages on, a candidate limit of 100 and a page size of 20. ## Pipelines and experiments An experiment variant can use a different recipe, and with it a different pipeline, so you can measure the effect of a stage. For example, run the rules stage off in one variant to see whether your rules help. See [Experiments](https://docs.neuronsearchlab.com/console/experiments.md). This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.