> ## Documentation Index
> Fetch the complete documentation index at: https://docs.extrascoop.app/llms.txt
> Use this file to discover all available pages before exploring further.

# Workspace Columns & AI Search

> How to organise real-time monitoring columns and leverage the AI-powered OmniCommand search engine.

The Newsroom Terminal workspace allows you to assemble tailored monitoring desks by configuring multi-column intelligence feeds and executing rapid AI-powered natural language queries or structured Boolean searches.

```mermaid theme={null}
graph TD
    A[Editor Query / Command] --> B[OmniCommand AI Gateway]
    B -->|Tier-0| C[Normalized Prompt Cache <1ms]
    B -->|Tier-1| D[Semantic Vector Cache]
    B -->|Inference| E[AI Natural Language Intent Compiler]
    C --> F[Structured Newsroom Query]
    D --> F
    E --> F
    F --> G[Real-Time Intelligence Data Grid]
    F --> H[AI Newsroom Briefing Synthesis]
```

## Managing Workspace Columns

You can configure multiple independent columns side-by-side to monitor different beats simultaneously:

<Steps>
  <Step title="Add a New Column">
    Click the **Add Column (+)** button in the workspace toolbar, or drag a filter preset directly onto the canvas.
  </Step>

  <Step title="Configure Column Filters">
    Click the filter icon in the column header to set specific criteria:

    * **Feed Scope:** Select *Latest* (chronological), *Top Stories* (rank-weighted), or *Hot* (velocity).
    * **Category:** Focus on specific beats such as *Crisis*, *Politics*, *Tech*, *Crime*, or *World*.
    * **Geographic Region:** Filter by city, borough, or country.
  </Step>

  <Step title="Reorder and Save Layout">
    Drag columns horizontally to reorder your workspace. The terminal automatically persists your layout locally across sessions.
  </Step>
</Steps>

## OmniCommand AI Search Engine

The **OmniCommand Bar** at the top of the terminal acts as an AI-powered intelligence compiler that translates complex editorial questions into high-speed database queries and synthesized briefings.

### How the AI Reduces Latency

To deliver sub-second search responses across massive media archives, the OmniCommand architecture uses a multi-tier caching and compilation pipeline:

```mermaid theme={null}
graph LR
    Query[Editor Prompt] --> Tier0[Tier-0: In-Memory Edge Cache]
    Tier0 -->|Hit ~0ms| Exec[Fast Database Execution]
    Tier0 -->|Miss| Tier1[Tier-1: Semantic Vector Cache]
    Tier1 -->|Similarity Match| Exec
    Tier1 -->|Miss| LLM[AI Intent Compiler]
    LLM --> Exec
```

1. **Tier-0 Edge In-Memory Cache (Sub-1ms):** Frequently requested newsroom queries and normalized prompt patterns are resolved instantly at the edge.
2. **Tier-1 Semantic Vector Cache:** Converts incoming queries into vector embeddings and matches semantically identical requests using cosine distance indexing without invoking LLM inference repeatedly.
3. **AI Natural Language Intent Compiler:** For novel prompts, a specialized language model extracts keywords, category beats, credibility requirements, media format preferences (video vs still photography), geographic bounds, and timeframes (e.g. 24h, 7d, 30d).
4. **Deterministic Client Fallback:** If offline or experiencing connection drops, a local heuristic compiler processes keywords instantly without interruption.

### AI Newsroom Briefing Synthesis

When you run a search, the OmniCommand engine can also generate an automated **AI Newsroom Briefing**—a concise 2–3 sentence factual summary synthesizing key breaking developments across all retrieved dispatches to accelerate editorial decision-making.

***

## Boolean Query Syntax

In addition to conversational AI prompts, editors can execute precise Boolean queries directly in the OmniCommand Bar:

| Operator            | Syntax Example           | Behaviour                                                                         |
| :------------------ | :----------------------- | :-------------------------------------------------------------------------------- |
| **AND**             | `parliament AND protest` | Matches dispatches containing both terms.                                         |
| **OR**              | `fire OR explosion`      | Matches dispatches containing either term.                                        |
| **NOT**             | `aviation NOT military`  | Excludes dispatches containing the specified term.                                |
| **Trust Threshold** | `trust:>85`              | Restricts results to dispatches with a Trust Score exceeding the specified level. |
| **Hashtags**        | `#Westminster`           | Filters by indexed community or newsroom hashtags.                                |

***

## Sonar Acoustic Radar

To ensure editors never miss breaking developments while multitasking across screens, the terminal includes a synthesised acoustic radar:

* **Radar Pings:** Emits a subtle acoustic chime whenever a newly published dispatch matching your column criteria hits the wire.
* **Custom Sensitivity:** Configure audio frequency, alert volume, or mute specific categories in **Terminal Settings > Audio**.