Skip to content

anzal1/quicky-wiki

Repository files navigation

⚡ Quicky Wiki

Turn any collection of documents into a living, confidence-scored knowledge base — powered by LLMs.

Quicky Wiki extracts claims from your sources, tracks how confident each claim is, watches for contradictions, and gives you a visual dashboard to explore everything. Think of it as a personal Wikipedia that actually tells you what it's unsure about.

npm version License: MIT Node.js >= 20


30-Second Demo

npx quicky-wiki init --name "My Research"   # auto-detects your API keys
qw ingest paper.pdf                          # extract claims from a paper
qw ingest https://arxiv.org/abs/2401.12345   # or a URL
qw serve                                     # open the dashboard

That's it. Open http://localhost:3737 and you've got:

  • A knowledge graph you can zoom into, hover over, and click through
  • Ask Wiki — chat with your knowledge base, get answers with confidence scores and citations
  • Claims, pages, timeline, and health views

What Makes It Different

Feature Traditional Wiki Quicky Wiki
Who writes it? You LLM extracts claims from your sources
Confidence? ❌ Everything looks equally true ✅ Every claim has a confidence score
Contradictions? Hidden in edit history Surfaced automatically
Temporal tracking? ✅ Claims strengthen, weaken, and decay over time
Knowledge gaps? Unknown unknowns Discovered and suggested
Multi-format output? Just a wiki Wiki, slides, flashcards, graph, timeline

Install

npm install -g quicky-wiki

The package exposes three equivalent commands: quicky-wiki, qw, and create-quicky-wiki (all run the same CLI). Use whichever fits your muscle memory or tooling.

Or use directly with npx:

npx quicky-wiki init

From a git clone of this repo, run npm run build before npx quicky-wiki …, node dist/cli.js …, or npm start, because the published CLI is the compiled dist/ bundle.

Requirements

  • Node.js >= 20
  • An API key from one of these providers:
Provider Env Variable Example Model
Google Gemini GOOGLE_API_KEY gemini-2.0-flash
OpenAI OPENAI_API_KEY gpt-4o
Anthropic ANTHROPIC_API_KEY claude-sonnet-4-20250514
Ollama (local, no key) llama3
Any OpenAI-compatible Custom env var + --base-url Groq, Together, vLLM, LM Studio

qw init auto-detects whichever API key you have set — no config needed.

Usage

Initialize

mkdir my-wiki && cd my-wiki
qw init --name "My Research"

# Or specify a provider explicitly
qw init --provider gemini --model gemini-2.0-flash
qw init --provider openai --model gpt-4o

# OpenAI-compatible (Groq, Together, etc.)
qw init --provider openai-compatible \
  --base-url https://api.groq.com/openai/v1 \
  --api-key-env GROQ_API_KEY

Ingest Sources

# Local files (markdown, text, PDF)
qw ingest paper.md --type paper --quality peer-reviewed
qw ingest notes.md --type note

# URLs (fetches and extracts automatically)
qw ingest https://example.com/article

# Batch ingest a directory
qw ingest raw/

Query Your Knowledge

qw query "What are the key approaches to reinforcement learning?"
qw query "How do transformers compare to RNNs?"

Explore with the Dashboard

qw serve                 # http://localhost:3737
qw serve --port 8080     # custom port

The dashboard includes:

  • Overview — Stats at a glance
  • Knowledge Graph — Interactive canvas visualization (Obsidian-inspired dark theme, hover to unfold connections)
  • Claims — Browse all extracted claims with confidence scores
  • Pages — Wiki pages compiled from your claims (optional entity kind and metadata when configured)
  • Timeline — Temporal view of knowledge events
  • Health — Knowledge integrity: stale claims, contradictions, gaps
  • Ask Wiki — Chat with your knowledge base

When you open a page, the slideout shows Linked pages (graph neighbors) and the rendered wiki markdown from wiki/. Inline Obsidian-style wikilinks work in that preview: [[Page Title]] and [[label|Page Title]] open the matching page by title. The dashboard script is embedded in the built CLI, so after changing TypeScript sources, run npm run build and restart serve to see UI updates.

Knowledge Health

qw lint                        # check for issues
qw metabolism --report         # full health report
qw metabolism --decay          # apply confidence decay over time
qw metabolism --resurface      # find stale claims worth revisiting
qw metabolism --redteam        # challenge high-confidence claims

Compile to Other Formats

qw compile markdown            # Obsidian-compatible wiki pages
qw compile slides --topic X    # Marp slide deck
qw compile anki                # flashcards
qw compile graph               # D3 knowledge graph
qw compile timeline            # temporal visualization

Discover New Directions

qw discover --mode gaps          # what's missing?
qw discover --mode horizon       # frontier topics
qw discover --mode bridges       # connections between distant concepts
qw discover --mode contradictions # conflicting claims

How It Works

Source Document
     ↓
LLM Extraction → Claims (with confidence scores)
     ↓                        ↓
Knowledge Graph (SQLite)    Epistemic Events (temporal log)
     ↓
Compiled Outputs: Wiki pages, slides, flashcards, graph, timeline
     ↓
Dashboard (interactive visualization + chat)

Key Concepts

  • Claim — An atomic, verifiable statement extracted from a source. Has a confidence score, provenance, and dependency chain.
  • Epistemic Event — A change in belief: created, reinforced, challenged, weakened, superseded, or resolved.
  • Knowledge Diff — When you ingest a new source, you see what's new, reinforced, challenged, and what gaps were found.
  • Cascade — When a foundational claim is challenged, confidence changes propagate through dependent claims.
  • Metabolism — Active maintenance: decay, resurfacing, red-teaming.

MCP Server

Quicky Wiki includes a built-in Model Context Protocol server for integration with AI agents:

qw mcp                          # stdio mode (for Claude Desktop, etc.)
qw mcp --http --port 3000       # HTTP mode

Tools include querying, search (full-text on the graph), listing pages and claims, ingestion, and health reporting. Pages can carry an entity kind and metadata (see config / compiled wiki frontmatter). MCP adds list_entities (filter by kind and metadata) and update_entity_metadata for merging metadata without re-ingesting; list_pages and ingest_file accept optional kind-related parameters. Use each tool’s schema in your MCP client for full argument lists.

Project Structure

my-wiki/
├── .quicky/
│   ├── config.yaml          # LLM provider, model, wiki name
│   └── graph.sqlite         # knowledge graph (claims, sources, events)
├── raw/                     # your source documents (immutable)
└── wiki/                    # compiled output (Obsidian-compatible markdown)

Development

git clone <repo-url> quicky-wiki && cd quicky-wiki
npm install
npm run build # required before serve / dist-based CLI picks up TS changes
npm run typecheck      # optional

License

MIT

About

⚡ LLM-powered knowledge compiler — turn documents into a confidence-scored, temporal knowledge base with interactive dashboard

Resources

License

Stars

Watchers

Forks

Packages

 
 
 

Contributors