Core Concepts

Knowledge Filtering

Knowlume is built around filtering, not only summarization.

Search is useful when you already know what you are looking for. Summaries are useful when you want a short overview of an entire source. Knowlume is useful when you do not yet know which parts of a source are valuable and need to inspect the source by quality signals.

The core question is:

Which parts of this source are worth saving into a knowledge base?

The answer is usually not the whole source. In a long podcast or lecture, only a few fragments may be reproducible, non-obvious, or source-rich enough to keep.

Project

A project is a workspace boundary. It groups sources, fragments, selections, and exports.

Use projects for:

• a research topic;
• a customer or client;
• a product area;
• a course or learning plan;
• a temporary investigation.

Every user has an Inbox project for unsorted work.

Source

A source is the original material submitted to Knowlume.

Supported source types include:

• website or article URL;
• YouTube URL;
• pasted text;
• file-based inputs where enabled.

Sources move through lifecycle states: queued, processing, waiting, succeeded, or failed. The public API uses the same external status vocabulary; see Integrations And API.

Processing

Processing is the work Knowlume performs after you add a source or request an export.

For a source, processing usually means:

1. fetch or receive source content;
2. extract readable text or transcript;
3. split content into chunks;
4. score and name fragments;
5. prepare fragments for review.

For an export, processing turns the selected fragment set into a Markdown artifact.

Fragment

A fragment is a distilled unit of knowledge. It is closer to a reusable note or micro-memory than to a generic summary paragraph.

Good fragments are:

• small enough to review quickly;
• large enough to preserve context;
• tied to a source;
• written as reusable Markdown;
• scored so they can be filtered.

Fragments are the main product output. The UI is built around reviewing, filtering, selecting, and exporting them. See Fragments, Scores, And Export.

Scores

Knowlume scores fragments on three 0-10 axes:

Score	Meaning
Reproducibility	Can the claim or method be checked, repeated, or operationalized?
Originality	Was the fragment's real effect hard to predict from its cause and context?
Sourcesness	Does the fragment explicitly mention books, articles, tools, services, websites, URLs, or other external sources?

Public API filters use inclusive integer ranges for these scores.

High reproducibility and high originality together are usually the strongest signal for a personal knowledge base: the fragment is both usable and worth saving.

Selection

A selection is the set of fragments the user wants to keep for export.

Selection is intentionally human-in-the-loop. Knowlume can rank and filter candidates, but the user decides what belongs in the final knowledge pack.

When reviewing a source, use scores to find candidates quickly, then open the fragment detail before saving anything important.

Export

An export converts selected fragments into Markdown.

Exports can be downloaded as files. They can also be sent through an Integration Webhook when the user configures webhook delivery.

Export artifacts are served from:

https://files.knowlu.me/knowlume-artifacts/...

Integration Surfaces

Knowlume has three main surfaces:

Surface	Audience	Base
Web app	Human users	https://app.knowlu.me
Public API	Automated integrations	https://app.knowlu.me/api/public/v1
MCP	Agent clients	https://app.knowlu.me/mcp/

For authentication, examples, and integration boundaries, continue with Integrations And API.