Overview

The wiki-base for the Latent Space community.

What Is This?

Latent Space Wiki-Base is a shared knowledge system for the Latent Space community. It combines markdown documentation with a simple relational database (SQLite). The documentation is living, breathing, evolving over time as content is ingested, enriched by AI, and explored by both humans and agents.

The core idea is externalised context. Rather than knowledge living in people's heads or inside agent context windows, it lives in one shared database that community members, contributors, and AI agents continuously read from and write to.

                    ┌─────────────────────┐
                    │                     │
    Humans ────────>│     Wiki-Base       │<──────── Agents
      ^             │   (Turso SQLite)    │             ^
      │             │                     │             │
      │             └─────────────────────┘             │
      │                    │      │                     │
      └────────────────────┘      └─────────────────────┘
         read / write via              read / write via
         web app + Discord             MCP tools + bot tools

The documentation, the database, and the AI agent layer are the same thing.

These docs describe how it all works. Each page covers a layer of the system.

One Repo, Two Deploy Targets

Path	What it is	Deployed on
latent-space-hub	Web app, ingestion pipeline, MCP server, evals	Vercel
`apps/bots/slop/`	Discord bot runtime (Slop), agentic tool-calling, member memory	Railway

They share one Turso database. The hub writes to it (ingestion, web UI, API). Slop reads from it with direct Turso queries and writes member/event data back.

Architecture

Four cloud services, no self-hosted infrastructure.

Service	Role
Vercel	Web app (read-only) + hourly ingestion cron + Discord announcements
Turso	Cloud SQLite, single shared database for web app, bot, and MCP
Railway	Slop bot, always-on Discord gateway with agentic tool-calling
OpenRouter	LLM routing for Slop (currently Claude Sonnet 4.6, swappable)

The Database

Metric	Count
Total nodes	~4,100+
Content nodes	~530 (podcasts, articles, AINews, workshops, paper clubs, builders clubs)
People & entity nodes	~3,400 (guests, organizations, topics)
Event nodes	~50 (scheduled paper club and builders club sessions)
Edges	~8,100
Chunks with embeddings	~36,000
Coverage	June 2023 to present, continuously updated

Web App

Next.js 15 app deployed on Vercel. The primary interface for browsing and managing the database.

Dashboard

Landing page with stats and 8 category cards. Each card shows node count and 3 preview items.

Search

Cmd+K global search across titles, descriptions, and content. Hybrid search: FTS5 + vector + Reciprocal Rank Fusion.

Fixed left panel with all 8 categories (icon + count badge), Quick Add input for pasting any URL or text to ingest, plus search, skills, evals, and docs access.

Tech Stack

Layer	Technology
Framework	Next.js 15 + TypeScript + Tailwind CSS
Database	Turso (cloud SQLite via `@libsql/client`)
Search	Turso native vector search (F32_BLOB + `vector_top_k`) + FTS5 + hybrid RRF
Embeddings	OpenAI `text-embedding-3-small` (1536d)
AI	OpenAI GPT-4.1-mini (entity extraction, descriptions)
MCP	Model Context Protocol server, `npx latent-space-hub-mcp`
Bot	Discord.js + OpenRouter, Claude Sonnet 4.6

Origin

Forked from RA-H, a local-first personal wiki-base. Heavily modified for community use. If you want to build something like this for your own personal context or for an enterprise installation, RA-H is completely open source.