Retriever Stages

Build powerful search pipelines by composing modular stages. Each stage transforms, filters, ranks, or enriches your documents.

Learn more in docs

Apply

Api Call

Enrich documents with external API calls

api_call

Apply

Json Transform

Transform documents using Jinja2 templates

json_transform

Apply

External Web Search

Search the web using Exa AI-native search

external_web_search

Apply

Document Enrich

Join and enrich documents with data from another collection

document_enrich

Apply

Web Scrape

Scrape and extract content from web pages

web_scrape

Feature Search

Search collections using multimodal embeddings

feature_search

Keyword Search

Full-text keyword search with BM25 ranking

keyword_search

Filter

Attribute Filter

Filter documents by metadata attributes

attribute_filter

Filter

Llm Filter

Filter documents using LLM-based evaluation

llm_filter

Filter

Score Filter

Filter documents by relevance score threshold

score_filter

Rank

Llm Rerank

Rerank documents using LLM-based relevance scoring

llm_rerank

Rank

Cross Encoder Rerank

Rerank documents using cross-encoder models

cross_encoder_rerank

Page 1 of 2

How Retriever Pipelines Work

Combine stages to build sophisticated search and retrieval pipelines. Each stage type serves a specific purpose in the data flow.

1. Search

Retrieve documents from collections using semantic search, keyword search, or hybrid approaches.

2. Filter

Remove documents that don't match criteria using attribute filters, score thresholds, or LLM evaluation.

3. Rank

Reorder documents by relevance using sorting, cross-encoders, or LLM-based reranking.

4. Generate

Create responses from retrieved documents using RAG, summarization, or custom LLM generation.