Skip to main content

GAIA

v0.1.1 Verified

The flagship GAIA agent — conversation, document Q&A, data analysis, and web research, with persistent memory and add-on skills

  • general
  • chat
  • rag
  • memory
  • skills
  • research

Install

Python package · 116.4 MB
gaia agent install gaia

Recommended — installs into your GAIA app and registers the agent automatically.

A local model must be running first: gaia init then lemonade-server serve.

Open in GAIA Docs & guides

About GAIA

@amd-gaia/gaia

One command gets you a working GAIA:

npx @amd-gaia/gaia

That fetches the two binaries GAIA needs — the agent sidecar and the terminal UI — verifies both against a checksum manifest that ships inside this package, and drops you into the terminal UI. No Python to install, no repo to clone, no build step. Everything runs locally on your machine; nothing you type or index leaves it.

The terminal UI you get here is the published terminal-hub component — the exact same binary a full GAIA install runs as gaia tui, not a separate build. So however you arrive at the terminal UI, it behaves identically.

What it actually does

  1. Resolves your platform — win32-x64, darwin-arm64, darwin-x64, linux-x64 (plus linux-arm64 / win32-arm64 for the terminal UI).
  2. Reads binaries.lock.json, the checksum manifest published with this exact package version. It records, per binary, which hub lane it comes from and what it must hash to.
  3. Downloads and SHA-256 verifies both binaries. A hash that does not match is a hard failure — the download is deleted and the run stops. There is no "continue anyway" path and no unverified fallback.
  4. Launches the terminal UI, which brings up the GAIA daemon and the agent sidecar and hands you the chat. Its exit code becomes ours.

Requirements

  • Node.js 18+ (for the built-in fetch).
  • Lemonade Server running locally — it hosts the model the agent thinks with. GAIA tells you if it isn't up.
  • The gaia Python CLI on PATH for the daemon the terminal UI starts. Install it with curl -fsSL https://amd-gaia.ai/install.sh | sh (Windows: irm https://amd-gaia.ai/install.ps1 | iex).

Supported platforms

The terminal UI is Go and cross-compiles everywhere. The agent sidecar is a frozen Python build, produced on the machine it targets, and has no arm64 Linux or arm64 Windows build. On those two platforms npx @amd-gaia/gaia stops with an error naming your platform and the ones that do work — it will not start a UI with no agent behind it.

Platform keyAgent sidecarTerminal UI
win32-x64✅✅
darwin-arm64✅✅
darwin-x64✅✅
linux-x64✅✅
linux-arm64—✅
win32-arm64—✅

npx @amd-gaia/gaia version prints this matrix, plus the version and source URL of each binary, for the release you have installed.

Commands

gaia [run] [options] [-- <tui args>]   Fetch + verify both binaries, then launch the TUI
gaia fetch [options]                   Download + verify only; print JSON and exit
gaia serve [options]                   Run the agent sidecar alone (REST API, no TUI)
gaia version                           Print the lock manifest and this host's platform
gaia help                              Show help

Anything after a bare -- goes to the terminal UI untouched:

npx @amd-gaia/gaia -- --debug

Common options:

FlagMeaning
--base-url <url>Override the download base URL from binaries.lock.json
--cache-dir <dir>Where to cache the terminal UI binary
--sidecar-dir <dir>Where to install the agent sidecar (default ~/.gaia/agents/gaia)
--platform <key>Fetch for another platform (fetch only)
--forceRe-download even when a verified binary is already cached
--port <n>Sidecar bind port for serve (default 8141)

Set DEBUG=gaia for download, spawn, and sidecar output on stderr. Diagnostics never touch stdout, which the terminal UI owns.

Where things land

WhatPath
Agent sidecar~/.gaia/agents/gaia/gaia-agent[.exe]
Terminal UI~/.gaia/npm-cache/gaia-<version>/gaia-tui[.exe]

The sidecar goes into the GAIA daemon's own cache directory on purpose: the daemon is what spawns and supervises it, and it does its own SHA-256 check on the way. By putting an already-verified binary there we save a second download rather than racing one.

The terminal UI is installed as gaia-tui, never as gaia — a file named gaia in a cache directory would shadow the gaia shim npm puts on your PATH.

Ports

ServicePort
Agent sidecar8141 on 127.0.0.1
GAIA daemonassigned at start, recorded in ~/.gaia/host/instance.json

Port 4001 is reserved repo-wide and is refused with an error if you pass it. Both services bind loopback only — this agent speaks for your documents and memory and has no business on a LAN interface.

Running the sidecar on its own

gaia serve skips the terminal UI and gives you the REST surface directly, for integrating GAIA into your own app:

npx @amd-gaia/gaia serve --port 8141
curl http://127.0.0.1:8141/health

It waits for GET /health, checks the contract version, and tears the whole process tree down on Ctrl+C. See SPEC.md for the endpoints.

Programmatic use

import { randomUUID } from "node:crypto";
import { fetchAll, startSidecar, shutdown } from "@amd-gaia/gaia";

const { sidecar } = await fetchAll();               // both binaries, SHA-256 verified
const proc = await startSidecar({ binaryPath: sidecar.binaryPath });

const sessionId = randomUUID(); // reuse across the whole conversation, see below

const res = await fetch(`${proc.baseUrl}/v1/gaia/query`, {
  method: "POST",
  headers: { "content-type": "application/json" },
  body: JSON.stringify({
    query: "summarize my notes",
    run_id: randomUUID(),
    session_id: sessionId,
    context: [],
  }),
});

await shutdown(proc);

/v1/gaia/query streams Server-Sent Events terminated by exactly one final or error. fetchAll() also returns the TUI's path if you would rather launch that.

Reuse the same session_id for every turn in a conversation. It is what lets a document you had it index, or a skill you had it load, survive to the next question — drop it (or mint a new one per call) and the agent still answers, but it forgets everything from the previous turn. See SPEC.md §5.2 for the retry and eviction behavior.

Every failure throws a typed error (IntegrityError, PlatformError, HealthTimeoutError, VersionMismatchError, BinaryNotFoundError) with a message that names what failed and what to do about it.

Where the binaries come from

The two binaries ship from two different places, and binaries.lock.json records a version and a source URL for each:

BinaryPublished asBuilt by
Agent sidecarthe gaia agent, at this package's versionthis package's release
Terminal UIthe terminal-hub component, at its own versionthe core GAIA release

The terminal UI is consumed, not rebuilt. It is byte-for-byte the gaia tui binary a core install ships, so there is no second copy that could lag behind or behave differently — which is the entire reason it is sourced this way.

Integrity

binaries.lock.json is the single source of truth for what gets downloaded and what it must hash to. The release pipeline regenerates it from the artifacts actually being served — the sidecars it just published, and the terminal-hub artifacts it downloaded and cross-checked against the hub's own recorded hashes.

Between releases the lock carries PENDING-replace-with-real-sha256 placeholders. A placeholder blocks the fetch outright — before any network call — so an unverifiable binary can never be downloaded, let alone executed. If you need to run against a locally built binary, build it yourself and point the lifecycle helpers at it directly; the fetcher will not be talked into it.

  • Guide: <https://amd-gaia.ai/docs/guides/gaia>
  • Technical reference: SPEC.md
  • Changes: CHANGELOG.md
  • Issues: <https://github.com/amd/gaia/issues>

MIT licensed. © 2024-2026 Advanced Micro Devices, Inc.