The AI Building Blocks
See Them Work

Four technologies. Live demos. No slides, no hand-waving — real inference, real parsing, real chains. Pick a building block or see them work together.

Live Inference Cloudflare Edge Open Source Stack

01 · Try It

Interactive Demos

⚡

Cloudflare AI

Type a prompt, pick a model, get a response from the edge. Streaming text generation, embeddings, and image classification — all serverless.

Live Workers AI

📄

LiteParse

See why layout-preserving extraction beats naive text dumps. Spatial vs naive parsing side by side, with bounding box visualization.

Live Document AI

🔗

LangChain

Build a chain, run a RAG pipeline, and compare raw API calls vs orchestrated AI — all with live inference from the edge.

Live Orchestration

🦙

llama.cpp

The honest comparison tool. Same prompt across quantisation levels, real benchmarks, and a hardware calculator that tells you what you actually need.

Coming Soon Local Inference

🔄

See All Four Work Together

Upload a document → parse it → embed chunks → store in a vector DB → ask questions → get answers with sources. The full RAG pipeline, live.

Coming Soon Full Pipeline

02 · When to Use What

The Decision Matrix

Every tool has a sweet spot. This is the honest guide to picking the right one.

Scenario	Use This	Why
Quick text generation, zero setup	Cloudflare AI	Pay-per-use, sub-second from the edge, no GPUs to manage
Parse documents for AI pipelines	LiteParse	Local, fast, spatial-aware — preserves layout LLMs can read
Multi-step AI workflow with retries	LangChain	Chain orchestration, memory, fallbacks, and observability built in
Air-gapped or offline inference	llama.cpp	Runs on bare metal, zero network dependency, full privacy
Full RAG pipeline	All Four →	Parse → Embed → Store → Retrieve → Generate
Prototype in an afternoon	Cloudflare AI + LangChain	Fastest path from idea to working demo
Privacy-sensitive data	LiteParse + llama.cpp	Everything on-premise, zero data leaves the building