Tools
Daniel Khanin*

Ulan-Bator Angarsk Almaty Tomsk Moscow Urmala Barcelona

Tools

loOom

tools

A graph explorer for structured data in the terminal

loOom is a TUI app that builds a knowledge graph from your CSV, ECSV, and Markdown files. Written in C with ncurses: a single binary, instant launch, works offline. Your data never leaves your machine.

You point it at your documents, loOom extracts entities (people, goals, teams, metrics) and automatically builds connections between them. Instead of reading dozens of files you see a map: who is connected to what, where data intersects, what dependency chains exist.

Installation

macOS (Intel + Apple Silicon):

$ tar xzf looom-0.2-macos.tar.gz $ cp looom-0.2-macos/looom /usr/local/bin/

Quick start

$ cd ~/Documents/my-data $ looom

loOom opens in the current directory. The left side shows the file browser and entity list; the right side shows an interactive graph.

  1. Navigate to a folder with files, press Enter to add them to the project
  2. Press X to extract entities from all files
  3. The graph is built — explore it

File browser

The upper-left panel provides file system navigation. Browse folders, add files, extract entities. Supported formats: CSV, ECSV, Markdown, XLSX.

  • Enter on a folder — open it
  • Enter on a file — add to project
  • s — show/hide file on graph
  • x — extract entities from file
  • X — extract from all files at once
  • v — view Markdown with syntax highlighting

Three file states: ○ not added, ● added and visible, ◔ added but hidden on graph.

Entity tree

The lower-left panel lists all extracted entities grouped by type. First level shows types with counts, second level shows individual entities.

  • s on a type — show/hide entire type on graph
  • Enter — expand type or open entity detail
  • F — focus graph on selected entity

Knowledge graph

An interactive force-directed graph of all entities and their connections. Vim-style navigation: h/j/k/l to move between nodes, H/J/K/L to pan, +/- to zoom, f for fullscreen.

Two display modes

Press m to switch:

  • All — files and entities with all connections
  • Entities — entities only with semantic relations (no files, no contains/found_in)

Hop depth

Press 1–9 to show only nodes within N hops of the selected node. * to show all. Instant switching with no graph recalculation.

Focus

Press F on any node — the graph redraws showing only the 2-hop neighborhood. Instead of 1000+ nodes you see 10–30 relevant ones. The central node is drawn in a frame above the rest.

Bridges between files

Press B to highlight entities that appear in two or more files. These are the connection points between your disparate data. Yellow nodes are bridges; everything else fades out.

Search

Press / from any panel. Results filter in real time by entity name and type. Enter focuses the graph on the result and opens detail view.

Detail view

Press Enter on an entity for a fullscreen view: type, name, source, attributes, and all connections (incoming and outgoing).

Filter by relation type

Press t to cycle through types: all → goal → person → team → metric → ... This lets you trace chains:

  1. Find person Sarah Chen → t → see her goals (G-101, G-105, G-108)
  2. Enter on G-101 → t → see metric Churn and product Payments
  3. Enter on Churn → t → see which files mention it

Entity extraction

loOom extracts entities using rules defined in .looom/entities.conf:

[goal] description: Quarterly business goal pattern: G-[0-9]+ color: accent [person] description: Team member or stakeholder csv_column: Owner csv_column: Lead color: accent2 [metric] description: Business KPI or metric pattern: MRR|ARR|ARPU|CAC|LTV|NPS|DAU|MAU csv_column: Metric color: warning

Three extraction mechanisms:

  • pattern — regular expression (POSIX), searched in every line of the file
  • csv_column — CSV column name; unique values become entities
  • ecsv_header — field in ECSV comment headers

Link rules

The [links] section in entities.conf defines semantic relationships:

[links] goal -> person: owned_by goal -> team: assigned_to goal -> metric: measures goal -> product: targets metric -> team: tracked_by

When entities of different types co-occur in the same CSV row, loOom automatically creates an edge between them. Flat tables become a knowledge graph.

AI agent

Press A to launch Claude Code CLI directly from loOom. If an entity is selected, Claude receives its full context: name, type, all connections.

Context from loOom: - Current entity: G-101 (type: goal) - Source: quarterly_goals.ecsv - person: Sarah Chen - team: Growth - metric: Churn - product: Payments

Use it to analyze entities, generate type configurations from your data, create reports, find anomalies. Press ? → Setup Claude and loOom will create a CLAUDE.md with instructions for the agent.

Markdown viewer

Press v on any .md file for a built-in viewer with highlighted headings, lists, blockquotes, code blocks, and tables. Works from the file browser, detail view, and column browser.

Themes

6 color themes: tokyonight-dark, tokyonight-light, nord-dark, nord-light, catppuccin-dark, catppuccin-light. Switch with T. All entity colors adapt automatically.

Help

Press ? to open the help panel on the right. It doesn't block your work — you can switch between panels freely. Sections: About, Hotkeys, Features, Usage examples, Prompts for Claude, Setup Claude, Activate License.

License and pricing

20€
One-time purchase. All updates within the major version included.

Trial: 2 days, up to 5 files and 50 entities. All features available.

License key is bound to your email and major version. Activate via ? → Activate License.

System requirements

  • macOS 10.15+ (Catalina or later)
  • Terminal with UTF-8 and true-color support
  • Intel or Apple Silicon

Recommended tools

  • vim — file editor (or set your $EDITOR)
  • csvview — CSV/ECSV viewer for the terminal — github.com/daniil-khanin/csvview
  • claude — AI agent — npm install -g @anthropic-ai/claude-code

© 2026 Daniil Khanin. All rights reserved. daniil@khanin.info