> For the complete documentation index, see [llms.txt](https://docs.atlas-ai.org/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.atlas-ai.org/features/intelligent-extraction.md).

# Intelligent Extraction

## Beyond Simple Storage

Most tools just store your content. ATLAS **understands** it.

When you add content to ATLAS, AI models analyze every piece to extract structured knowledge that's searchable, connectable, and actionable.

***

## What Gets Extracted

### 📌 Concepts

**Key ideas with importance scores**

Every piece of content yields concepts — the fundamental ideas it contains. Each concept includes:

| Field       | Description                   |
| ----------- | ----------------------------- |
| Name        | Clear, descriptive identifier |
| Description | What this concept means       |
| Importance  | Score from 0.0 to 1.0         |
| Category    | Domain classification         |
| Source      | Where it came from            |

**Example:**

```
Name: "Antifragility"
Description: "Systems that gain from disorder and volatility"
Importance: 0.92
Category: Systems Thinking
Source: Twitter bookmark on Taleb thread
```

### 💡 Insights

**Actionable learnings and observations**

Not just facts, but implications. What should you remember? What matters?

**Example:**

```
"Most productivity systems fail because they optimize 
for capture rather than retrieval. Design for finding, 
not filing."
```

### ✅ Actions

**Things to do, research, or explore**

Your content often implies next steps. ATLAS captures them:

**Example:**

```
- Research: "Look into Zettelkasten methodology"
- Build: "Create a personal knowledge API"  
- Connect: "Reach out to author for interview"
```

### 📝 Quotes

**Memorable passages worth preserving**

The exact words that captured an idea perfectly.

### 🧩 Frameworks

**Mental models and thinking tools**

Structured approaches to problems. Decision matrices. Evaluation criteria. Reusable thinking patterns.

***

## The Extraction Process

### Step 1: Content Analysis

LLM receives the full content with extraction prompts optimized for each content type (tweets vs. long articles vs. conversations).

### Step 2: Structured Output

AI returns JSON-structured extractions that map to ATLAS's schema. No fuzzy outputs — clean, queryable data.

### Step 3: Importance Scoring

Each concept gets an importance score based on:

* Relevance to your existing knowledge graph
* Novelty of the idea
* Actionability
* Connection potential

### Step 4: Deduplication

If a concept already exists in your graph, ATLAS merges the new reference rather than creating duplicates.

### Step 5: Relationship Mapping

New concepts automatically link to related existing concepts based on semantic similarity and co-occurrence.

***

## Importance Scoring

The 0.0 to 1.0 importance score is crucial for surfacing what matters.

### Scoring Factors

| Factor         | Weight | Description                       |
| -------------- | ------ | --------------------------------- |
| Novelty        | 25%    | Is this new to your graph?        |
| Density        | 25%    | How connected to other concepts?  |
| Actionability  | 20%    | Does it suggest concrete actions? |
| Source Quality | 15%    | Where did this come from?         |
| Recency        | 15%    | Is this timely?                   |

### What Scores Mean

* **0.9 - 1.0**: Core concepts, foundational ideas
* **0.7 - 0.9**: Important concepts, frequent references
* **0.5 - 0.7**: Useful supporting concepts
* **0.3 - 0.5**: Contextual, situational concepts
* **0.0 - 0.3**: Peripheral, low-signal concepts

***

## Current Stats

| Extracted Item    | Count   |
| ----------------- | ------- |
| **Concepts**      | 12,768+ |
| **Insights**      | 5,691+  |
| **Actions**       | 4,694+  |
| **Relationships** | 9,807+  |
| **Content Items** | 5,731+  |

***

## The Compounding Effect

Here's what makes intelligent extraction powerful:

### Month 1

You add your Twitter bookmarks. ATLAS extracts 500 concepts.

### Month 3

You add Apple Notes and GitHub stars. ATLAS extracts 1,000 more concepts — but 300 of them connect to existing ones, enriching both.

### Month 6

Your graph has 3,000 concepts with 8,000 relationships. New content immediately slots into context.

### Month 12

ATLAS knows your intellectual landscape. It can answer questions about your specific interests, connect disparate ideas, and surface forgotten insights at the right moment.

**Knowledge compounds. Storage doesn't.**


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.atlas-ai.org/features/intelligent-extraction.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.