vectorless-docs

Documentation site for Vectorless — document retrieval for the reasoning era. No chunking, no embeddings, no vector DB: Vectorless parses a document into a tree, an LLM agent navigates it with treewalk, and returns answers with citations.

Built with Fumadocs on Next.js (App Router). Deploys to docs.vectorless.store.

Develop

pnpm install
pnpm dev        # http://localhost:3000

Build

pnpm build
pnpm start

Content

Docs live in content/docs/**.mdx. Sidebar order and section grouping are controlled by the meta.json files. Sections:

• getting-started — positioning + quickstart
• concepts — tree retrieval, no-chunking model, treewalk, citations
• api — API reference (placeholder; will render from the engine OpenAPI spec)
• sdks — TypeScript, Python, Go
• self-hosting — Neon + Cloudflare R2 + Upstash QStash + Docker (placeholder)

Design

Dark theme primary. Geist Sans (body/UI) + Geist Mono (code) + Instrument Serif (display headlines). Blue→pink brand gradient (#1456F0 → #3B82F6 → #EA5EC1), subtle grid background texture. Brand tokens live in src/app/global.css; fonts in src/fonts wired via src/lib/fonts.ts.

Deploy (next step — needs Vercel + DNS auth)

Not yet deployed. To ship: import the repo into Vercel and point docs.vectorless.store at it (Spaceship DNS).