# Architecture



# Overview

Peam is organized around a build-time pipeline and a runtime query stack. The builder ingests your content through [index sources](/docs/indexing/sources), applies [filters](/docs/indexing/filters), and produces index artifacts that are written to [index stores](/docs/storage). At runtime, the search engine loads those artifacts to serve AI-powered answers to your client.

## High-level architecture

<div className="mx-auto w-full max-w-3xl">
  <Mermaid
    chart={`
flowchart
subgraph Runtime[Runtime]
  direction TB
  Client[Client UI / SDK]
  Server[Peam Server API]
  SearchEngine[Search engine]
  Retriever[Retrieval + reranking]
  LLM[LLM response]
  Client --> Server
  Server --> SearchEngine
  SearchEngine --> Retriever
  Retriever --> LLM
  LLM --> Client
end

subgraph Build_Time[Build-time]
direction TB
Sources[Docs, files, APIs]
IndexSources[Index sources]
Filters[Filters]
Indexers[Indexers]
Stores[Index stores]
Sources --> IndexSources
IndexSources --> Filters
Filters --> Indexers
Indexers --> Stores
end
`}
  />
</div>

## What each part does

* **Index sources** discover pages or content to index. See [Index sources](/docs/indexing/sources).
* **Filters** prune or refine candidates before indexing. See [Filters](/docs/indexing/filters).
* **Peam Builder** orchestrates ingestion, normalization, and indexing.
* **Indexers** parse and structure content, producing embeddings, metadata, and search artifacts.
* **Index stores** persist index artifacts for runtime search. See [Storage](/docs/storage).
* **Search engine** loads stored artifacts and powers retrieval.
* **Runtime server** handles user queries, retrieval, and answer composition.
* **Client** renders [AskAI](/docs/components/ask-ai) surfaces and sends messages to the runtime API.

Next, see [Data flow](/docs/core-concepts/data-flow) to understand how a single query moves through the system.


## Sitemap
[Overview of all docs pages](/sitemap.md)