# Data flow



# Overview

Peam uses a two-phase workflow: [indexing](/docs/indexing) at build time and querying at runtime. The builder creates artifacts once; the runtime reuses them for fast, consistent answers.

## Indexing flow

<Mermaid
  chart={`
sequenceDiagram
autonumber
participant Sources as Content Sources
participant Source as Index Sources
participant Filters as Filters
participant Indexer as Indexers
participant Store as Index Stores

Sources->>Source: Discover pages
Source->>Filters: Apply filters
Filters-->>Indexer: Filtered pages
Filters->>Indexer: Normalize + chunk
Indexer->>Indexer: Enrich metadata
Indexer->>Store: Write index
`}
/>

## Query flow

<Mermaid
  chart={`
sequenceDiagram
autonumber
participant Client as Client UI
participant Server as Peam Server
participant Search as Search Engine
participant Store as Index Stores
participant Retriever as Retrieval
participant LLM as LLM

Client->>Server: sendMessage()
Server->>Search: search(query)
Search->>Store: Load index artifacts
Store-->>Search: Index data
Search->>Retriever: Retrieve + rerank
Retriever-->>Server: Ranked context
Server->>LLM: Compose answer
LLM-->>Client: Streamed response
`}
/>

## Notes

* Index sources define where content is discovered from. See [Index sources](/docs/indexing/sources).
* Filters narrow the candidate set before indexing. See [Filters](/docs/indexing/filters).
* Index stores persist artifacts for runtime search. See [Storage](/docs/storage).
* Client surfaces stay thin: they render [AskAI](/docs/components/ask-ai) UI and send messages to the server.


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