Docs: Add Architecture page (#169)

## Description 📝

Adds a page for Architecture that's purposefully very broad.

If you take a peek at the categories in the sidebar, we enter with the
quickstart for end users and then build by talking about capabilities,
steering, and use cases.

This is the first page in the Core Concepts section and is meant to be
as high-level of an overview as possible. We could—and should—introduce
a section within this page that highlights differences between the data
and control planes. But, for now, this simply serves to illustrate the
various services that are running and how they work together to produce
the end-user experience.

An easy way to think of this: what's (relevant) in the root
`compose.yaml` and what does each service do?

## Quick Links 🚀


[Architecture](https://robdominguez-doc-2910-add-ar.promptql-docs.pages.dev/architecture/)

Author: robertjdominguez

docs/architecture.mdx

@@ -0,0 +1,74 @@
+---
+sidebar_position: 1.9
+sidebar_label: Architecture
+description:
+  "Learn about the high-level architecture of PromptQL and the importance of the semantic metadata layer in giving you
+  reliable and accurate answers to your questions."
+keywords:
+  - promptql
+  - architecture
+---
+
+# Architecture
+
+## Introduction
+
+Below, you'll find a functional overview of PromptQL's architecture and the key components that power it. At the heart
+of PromptQL is the **semantic metadata layer** — the intelligent foundation that makes natural language data queries
+possible.
+
+## High-level overview
+
+```mermaid
+
+flowchart TD
+    Client[Client] --> PlaygroundServer[Playground Server]
+    Metadata[Semantic Metadata] --> PlaygroundServer
+    PlaygroundServer --> LLM[LLM-of-choice]
+    LLM --> QueryPlan[Query Plan]
+    QueryPlan --> PlaygroundServer
+    PlaygroundServer --> Runtime[Python Runtime]
+    Runtime --> Engine[Distributed Query Engine]
+    Engine --> Connectors[Data Connectors]
+    Connectors --> DataSources[Data Sources]
+```
+
+## How it works
+
+### Request initiation
+
+When you ask PromptQL a question, the **Playground Server** receives your natural language query and combines it with
+your **semantic metadata** — the intelligent layer that describes your connected data sources, schemas, relationships,
+business logic, and access controls.
+
+This metadata layer is generated automatically when you connect a data source and evolves as you and your team use
+PromptQL.
+
+### Query plan generation
+
+The Playground Server sends both your question and the semantic metadata to your chosen LLM. **This semantic context is
+what transforms a general-purpose LLM into a data expert for your specific organization.** The metadata tells the LLM
+exactly what data exists, how it's structured, what different fields mean in your business context, and how tables
+relate to each other.
+
+With this rich context, the LLM generates a precise query plan that maps your natural language request to the right data
+operations.
+
+### Query execution
+
+The Playground Server passes the query plan to PromptQL's **Python runtime**, which executes it through the
+**distributed query engine**. The engine uses **data connectors** to fetch data from your **data sources**, respecting
+all access controls and mappings defined in your metadata.
+
+### Response delivery
+
+The retrieved data flows back through the same path: from data sources through connectors, engine, and runtime, where
+it's packaged and returned to you via the Playground Server.
+
+This architecture ensures that your LLM has complete context about your data landscape while maintaining security and
+performance through the distributed execution layer.
+
+## Next steps
+
+With a broad perspective of the different services powering a PromptQL application, we recommend learning more about the
+[semantic metadata layer](/data-modeling/overview.mdx) next.

src/theme/DocSidebar/categories.ts

@@ -12,7 +12,7 @@ export const CATEGORY_CONFIG: Record<string, CategoryConfig> = {
   },
   coreConcepts: {
     title: 'Core Concepts', 
-    directories: ['data-modeling', 'data-sources', 'business-logic', 'auth'],
+    directories: ['architecture', 'data-modeling', 'data-sources', 'business-logic', 'auth'],
     exactMatch: false,
   },
   buildingApps: {