SAGE — The People's AI

What is SAGE?

An AI that runs on your computer — for free, forever.

SAGE is a decentralized AI that lives on your machine. No cloud, no accounts, no subscription. Just install and start chatting.

But here's the magic: every person who runs SAGE makes it smarter for everyone. Your node learns from your conversations, encodes that knowledge into compact neural patterns, and shares it across the network.

Your raw data never leaves your machine. Only anonymous, compressed knowledge patterns flow between nodes. Privacy by design.

Think of it like BitTorrent, but for intelligence.

🆓

Free Forever

No API keys. No rate limits. No billing surprises. SAGE runs on your hardware — you already own everything you need.

🧠

Gets Smarter

Every node that joins the network contributes knowledge. The more people run SAGE, the smarter it gets — for everyone.

🔒

Your Data Stays Yours

Raw conversations never leave your machine. Only encrypted knowledge patterns are shared. PII filtering and local-only mode available.

How It Works

Four Steps to Collective Intelligence

From install to contributing to a global intelligence network — in under a minute.

📦

Install

One command. 30 seconds. Runs on any machine — even old laptops. No GPU required.

💬

Chat

Run sage chat and talk to your local AI. It learns from every conversation.

🔗

Connect

Run sage node start — your node joins the mesh. Knowledge begins flowing.

🌱

Grow

Every node makes the whole network smarter. Collective intelligence, emergent and unstoppable.

Why SAGE?

The AI You Deserve

Stop renting intelligence. Own it.

	SAGE	ChatGPT	Ollama	Cloud APIs
Free forever	✓	✗ $20/mo	✓	✗ pay/token
Runs locally	✓	✗	✓	✗
Gets smarter over time	✓ network	✗	✗	✗
No account needed	✓	✗	✓	✗
Privacy-first	✓	✗	✓	✗
Decentralized	✓	✗	✗	✗
No rate limits	✓	✗	✓	✗
OpenAI compatible	✓	is OpenAI	✓	✓

vs ChatGPT

Free, private, no account needed. SAGE gets smarter from the community — ChatGPT gets smarter from selling your data.

vs Ollama

Same local-first philosophy, but SAGE learns and shares knowledge across nodes. Your Ollama is an island. Your SAGE is a network.

vs Cloud APIs

No rate limits. No billing. No vendor lock-in. Drop-in OpenAI-compatible API. Switch in one line.

Under the Hood

Technical Deep Dive

For the nerds. You know who you are.

🧬

Grid-Based Knowledge — 256×256 Semantic Embedding Space

▼

SAGE's core is a 256×256 grid-based knowledge store with 38 channels per cell. Knowledge is encoded as 64-dimensional semantic embeddings distributed across the grid using semantic hashing. Attention-based retrieval finds relevant knowledge via cross-attention mechanisms, with feedback learning that improves results over time. Simple queries hit the grid directly (~73ms response). Complex queries seamlessly fall back to LLM when needed. Note: The NCA (Neural Cellular Automata) dynamics are aspirational — current implementation uses static knowledge channels with neighbor-averaged hidden states, not learned cellular automata rules.

🔄

Knowledge Sync — Encrypted Diffs, Not Your Conversations

▼

Nodes exchange compact knowledge diffs via a gossip protocol — not your raw conversations. These diffs are anonymous, compressed knowledge representations measured in kilobytes, not gigabytes. Peer discovery uses libp2p. Sync is lazy and runs in the background — never blocks your chat. Bootstrap peers now configurable via ~/.sage/config.toml for easy WAN connections.

⚡

Intelligent Query Router — Learns What You Ask

▼

SAGE v0.3.7 introduces a learning-based query router that replaces static heuristics. The router detects 12 query patterns (FactualLookup, Temporal, Comparative, Causal, Procedural, etc.) and learns which patterns work best with grid retrieval vs LLM fallback. Simple single-hop queries get 73ms responses from the grid. Complex multi-hop queries route directly to the model. The router tracks per-pattern accuracy and improves over time. Weights persist to disk and carry across sessions.

🔒

Privacy — PII Filtering, Differential Privacy, Local-Only Mode

▼

Raw data never leaves your machine. Knowledge is encoded into embeddings before any network sharing occurs. Built-in PII filtering strips personal information before encoding. Differential privacy adds calibrated noise to shared patterns. And if you want total isolation, local-only mode keeps everything on your machine with zero network traffic.

🔌

OpenAI Compatible — Drop-In Replacement for Any Tool

▼

SAGE exposes an OpenAI-compatible API on localhost. Set OPENAI_API_BASE=http://localhost:19176/v1 and any tool that speaks the OpenAI protocol — Continue, Cursor, Open WebUI, custom scripts — works with SAGE instantly. Zero code changes. No API key needed.

The Vision

What if AI belonged to everyone?

Every SAGE node contributes to a collective intelligence. The more people run it, the smarter it gets — for free. No corporation owns it. No paywall gates it. Just people, running AI, together.

✓ Phase 1

Local chat

✓ Phase 2

Network sync

Phase 3 — Active

Intelligent routing, bootstrap config

Phase 4

Trained NCA dynamics

The People's AI