LocalAI
LocalAI is the open-source AI engine. Run any model - LLMs, vision, voice, image, video - on any hardware. No GPU required.
- 50k
- Go
- MIT
- 50 min ago
:pushpin: Yaml Instructions Processor - Simply applies a cloud-init style yaml file to the system
Repo mudler/yip, read from GitHub 12 min ago. mudler has not joined Take My Repo.
mudler/yip is a public repo on GitHub by mudler. :pushpin: Yaml Instructions Processor - Simply applies a cloud-init style yaml file to the system. It declares the Apache-2.0 license on GitHub. It is written mainly in Go, has 117 stars and last changed on 2026-11-19. Nobody has reviewed it on Take My Repo yet, which says nothing about its quality. Read from GitHub on 2026-11-22.
In the maintainer's words. Shown from GitHub; Take My Repo does not keep a copy.
Applies a configuration described in YAML to the system. Distro-agnostic, cloud-init style, small scope, pluggable.
name: "Example"
stages:
default:
- name: "Write a file and run it"
files:
- path: /tmp/hello.sh
content: |
#!/bin/sh
echo "hello from yip"
permissions: 0755
commands:
- /tmp/hello.sh
$> yip -s default file.yaml
$> yip -s default https://example.com/file.yaml
$> cat file.yaml | yip -
Yip runs stages. A stage is a named list of steps, and each step is a map of fields that trigger built-in plugins. For example:
stages:
default:
- files:
- path: /tmp/bar
content: "hello"
permissions: 0644
commands:
- cat /tmp/bar
The step above triggers two plugins in the same step: files (writes /tmp/bar) and commands (runs cat). Within a single step plugins run in a fixed order — you do not need to split them across steps to guarantee ordering.
A YAML file can define multiple stages, and yip can be told which one to run with -s <stage>. Multiple files can be passed on the command line; their stages are merged before execution.
If any step fails yip exits non-zero, but it keeps running the remaining steps and files.
yip loads cloud-init style yamls and applies them in the system.
For example:
$> yip -s initramfs https://<yip.yaml> /path/to/disk <definition.yaml> ...
$> yip -s initramfs <yip.yaml> <yip2.yaml> ...
$> cat def.yaml | yip -
Usage:
yip [flags]
Flags:
-e, --executor string Executor which applies the config (default "default")
-h, --help help for yip
-s, --stage string Stage to apply (default "default")
Sources can be local paths, HTTP(S) URLs, or - for stdin.
A subset of the official cloud-config spec is understood. If a file starts with #cloud-config it is parsed as cloud-init and mapped to the yip boot stage:
#cloud-config
users:
- name: "bar"
passwd: "foo"
groups: "users"
ssh_authorized_keys:
- faaapploo
ssh_authorized_keys:
- asdd
runcmd:
- foo
hostname: "bar"
write_files:
- encoding: b64
content: CiMgVGhpcyBmaWxlIGNvbnRyb2xzIHRoZSBzdGF0ZSBvZiBTRUxpbnV4
path: /foo/bar
permissions: "0644"
owner: "bar"
Run with yip -s boot cloud-config.yaml.
Host data collected by sysinfo is exposed to templates in commands, files, and entity fields:
stages:
default:
- name: "echo hostname"
commands:
- echo "{{.Values.node.hostname}}"
name: "Test yip!"
Each step supports a set of predicates that decide whether it runs. Predicates are AND-ed: all of them must pass for the step to execute.
ifEvaluates the expression as a shell command. The step runs only when the command exits 0.
stages:
default:
- name: "debug boot only"
if: "cat /proc/cmdline | grep debug"
commands:
- echo "debug enabled"
if_filesRuns (or skips) the step based on file existence. Accepts three optional sub-lists:
any: at least one of the files exists.all: all of the files exist.none: none of the files exist.Combine them freely — the step runs when every provided sub-condition is satisfied.
stages:
default:
- name: "only on Linux hosts that never saw yip"
if_files:
any:
- /etc/os-release
all:
- /etc/passwd
none:
- /var/lib/yip/done
commands:
- touch /var/lib/yip/done
nodeRuns the step only on hosts whose hostname matches the pattern. Accepts a Go regexp.
stages:
default:
- node: "bastion.*"
commands:
- echo "runs on bastion hosts only"
only_os / only_os_versiononly_os matches the PRETTY_NAME field from /etc/os-release, and
only_os_version matches VERSION_ID. Both are compiled as Go regexps, which
are RE2: there is no lookahead,
and matching is case-sensitive and unanchored.
PRETTY_NAME is the human-readable string, so it is Ubuntu 24.04.1 LTS
rather than the ubuntu of ID. A pattern written against ID matches
nothing.
stages:
default:
- name: "ubuntu or opensuse-leap"
only_os: "Ubuntu.*|openSUSE Leap.*"
commands:
- echo hello
- name: "everything but ubuntu"
# RE2 has no negative lookahead, so exclude by listing what to include.
only_os: "Debian.*|Fedora.*|Alpine.*"
commands:
- echo hello
- name: "ubuntu 20.04 or 22.04"
only_os: "Ubuntu.*"
only_os_version: "20.04|22.04"
commands:
- echo hello
only_archMatches runtime.GOARCH (amd64, arm64, riscv64, ...) against a Go regexp.
stages:
default:
- name: "only on arm64"
only_arch: "arm64"
commands:
- echo "arm64!"
- name: "arm64 or amd64"
only_arch: "arm64|amd64"
commands:
- echo hello
only_service_managerRuns the step only when the host uses the given service manager. Supported values: systemd, openrc. Detection is done by probing for the corresponding binaries; a system where both are present is treated as ambiguous and the step is skipped.
stages:
default:
- name: "enable service under systemd only"
only_service_manager: "systemd"
systemctl:
enable:
- my-service
Below is a reference of all keys available in a step.
stages.<stageID>.[<stepN>].nameHuman-readable description of the step. Used only when printing progress to the console.
stages:
default:
- name: "Setup logging"
commands:
- echo "started"
stages.<stageID>.[<stepN>].afterDeclares that this step depends on other named steps. Yip uses this to build a dependency graph within a stage. The referenced names must match the name field of other steps in the same stage.
stages:
default:
- name: "write config"
files:
- path: /etc/myapp.conf
content: "key=value"
- name: "start service"
after:
- name: "write config"
commands:
- systemctl start myapp
stages.<stageID>.[<stepN>].directoriesCreates directories on disk. Runs before files so that file writes into fresh directories succeed in one step.
stages:
default:
- name: "Setup folders"
directories:
- path: "/etc/foo"
permissions: 0600
owner: 0
group: 0
stages.<stageID>.[<stepN>].filesWrites files to disk.
Supported encoding values: b64, base64, gz, gzip, gz+base64, gzip+base64, gz+b64, gzip+b64. Owner and group can be given as numeric uid/gid (owner, group) or as a string (ownerstring: "user:group" or "user").
stages:
default:
- files:
- path: /tmp/bar
content: |
#!/bin/sh
echo "test"
permissions: 0755
owner: 1000
group: 100
- path: /etc/motd
encoding: b64
content: SGVsbG8gd29ybGQK
ownerstring: "root:root"
stages.<stageID>.[<stepN>].downloadsDownloads files over HTTP(S) and writes them to disk. timeout is in seconds; 0 means no timeout.
stages:
default:
- downloads:
- path: /usr/local/bin/tool
url: "https://example.com/tool"
timeout: 30
permissions: 0755
ownerstring: "root:root"
stages.<stageID>.[<stepN>].gitClones a git repository into path. Supports HTTPS with username/password and SSH with a private key. branch_only: true fetches only the specified branch (single-branch checkout).
stages:
default:
- name: "Clone config repo"
git:
url: "https://github.com/example/config.git"
path: /opt/config
branch: "main"
branch_only: true
auth:
username: "gituser"
password: "gitpassword"
- name: "Clone via SSH key"
git:
url: "git@github.com:example/private.git"
path: /opt/private
branch: "main"
auth:
private_key: |
-----BEGIN OPENSSH PRIVATE KEY-----
...
-----END OPENSSH PRIVATE KEY-----
# optional; set to true to skip host key verification
insecure: true
stages.<stageID>.[<stepN>].usersAdds or modifies users. Each entry is keyed by username and takes the following fields. Every field is optional and of type string unless noted otherwise. When a user already exists, only the password is updated.
/home/<name>.github:<name> / gitlab:<name> form are resolved by fetching the user's public keys from the corresponding service.! in /etc/shadow)./etc/passwd is regenerated on every boot but /home is persisted, to keep file ownership stable.primary_group is empty), the primary group is created with this exact gid. If primary_group is set too, that group's existing gid is used and gid is ignored.start:count format. Yip creates or replaces the user's entry in /etc/subuid.start:count format. Yip creates or replaces the user's entry in /etc/subgid.stages:
default:
- name: "Setup users"
users:
bastion:
passwd: "strongpassword"
homedir: "/home/bastion"
shell: "/bin/bash"
groups:
- wheel
ssh_authorized_keys:
- github:mudler
svc-app:
uid: "12345"
subuid: "100000:65536"
subgid: "100000:65536"
system: true
shell: "/usr/sbin/nologin"
stages.<stageID>.[<stepN>].authorized_keysAdds SSH keys to the authorized_keys file of an existing user. Entries starting with github: or gitlab: are fetched from those services.
stages:
default:
- name: "Provision SSH keys"
authorized_keys:
mudler:
- github:mudler
- "ssh-rsa AAAA..."
root:
- "ssh-ed25519 AAAA..."
stages.<stageID>.[<stepN>].ensure_entitiesAdds a user or group in the entities format. Useful when you need full control over the /etc/passwd, /etc/shadow, or /etc/group row.
stages:
default:
- name: "Ensure system user"
ensure_entities:
- path: /etc/passwd
entity: |
kind: "user"
username: "foo"
password: "x"
uid: 1500
gid: 1500
info: "Foo service user"
homedir: "/home/foo"
shell: "/bin/bash"
stages.<stageID>.[<stepN>].delete_entitiesRemoves a user or group entity from the system.
stages:
default:
- name: "Drop legacy user"
delete_entities:
- path: /etc/passwd
entity: |
kind: "user"
username: "legacy"
stages.<stageID>.[<stepN>].hostnameSets the machine hostname on the running system, writes it to /etc/hostname, and updates /etc/hosts.
stages:
default:
- hostname: "node-01"
stages.<stageID>.[<stepN>].dnsWrites /etc/resolv.conf (or the file specified in path).
stages:
default:
- name: "Setup DNS"
dns:
nameservers:
- 8.8.8.8
- 1.1.1.1
search:
- example.com
options:
- "timeout:2"
path: "/etc/resolv.conf"
stages.<stageID>.[<stepN>].sysctlSets kernel parameters via /proc/sys/<key>, equivalent to sysctl -w.
stages:
default:
- name: "Kernel tuning"
sysctl:
net.ipv4.ip_forward: "1"
debug.exception-trace: "0"
stages.<stageID>.[<stepN>].modulesLoads kernel modules.
stages:
default:
- name: "GPU driver"
modules:
- nvidia
- nvidia_uvm
stages.<stageID>.[<stepN>].environmentWrites variables to /etc/environment (or the file specified in environment_file).
stages:
default:
- environment:
FOO: "bar"
HTTP_PROXY: "http://proxy:3128"
stages.<stageID>.[<stepN>].environment_fileOverrides the file written by environment. Use it to target a user-specific env file such as ~/.envrc instead of the system-wide /etc/environment.
stages:
default:
- environment_file: "/home/user/.envrc"
environment:
FOO: "bar"
stages.<stageID>.[<stepN>].timesyncdWrites a drop-in file at /etc/systemd/timesyncd.conf.d/10-yip.conf (under the [Time] section). Any key from timesyncd.conf can be set.
stages:
default:
- name: "Setup NTP"
systemctl:
enable:
- systemd-timesyncd
timesyncd:
NTP: "0.pool.ntp.org 1.pool.ntp.org"
FallbackNTP: ""
stages.<stageID>.[<stepN>].systemd_firstbootRuns systemd-firstboot with the given map. Keys map to --<key>=<value> arguments (keymap, locale, timezone, ...).
stages:
default:
- systemd_firstboot:
keymap: us
locale: en_US.UTF-8
timezone: Europe/Madrid
stages.<stageID>.[<stepN>].systemctlManages systemd unit state and optionally drops override files. All lists are optional.
.service is appended if missing). Required.override-yip.conf. .conf is appended if missing.Overrides are written even if the unit is not enabled or does not exist yet.
stages:
default:
- name: "Configure services"
systemctl:
enable:
- systemd-timesyncd
- cronie
mask:
- purge-kernels
disable:
- crond
start:
- cronie
overrides:
- service: "systemd-timesyncd"
name: "override-custom.conf"
content: |
[Service]
ExecStart=
ExecStart=/usr/lib/systemd/systemd-timesyncd
stages.<stageID>.[<stepN>].packagesInstalls, removes, refreshes, or upgrades packages using the host package manager. The distro is detected from /etc/os-release; supported managers: apt-get (Debian/Ubuntu), dnf (Fedora/RHEL/Rocky/Alma/Oracle/CentOS/openEuler), zypper (openSUSE/SLE), pacman (Arch), apk (Alpine).
Operation order is: refresh → upgrade → install → remove.
Each of the four operations is run up to three times. A package manager reports a mirror it could not reach as a plain command failure, and re-running the command makes it resolve the mirror list again, so a mirror that is briefly unreachable no longer costs the whole stage. The waits between the attempts are 5 and 15 seconds.
stages:
default:
- name: "Base packages"
packages:
refresh: true
upgrade: false
install:
- vim
- curl
remove:
- nano
stages.<stageID>.[<stepN>].package_pinsBest-effort version pinning applied before package installs. The exact mechanism depends on the detected package manager:
/etc/apt/preferences.d/99-yip-pins with Pin-Priority: 1001./etc/dnf/plugins/versionlock.conf + a versionlock.list. Full NEVRA is resolved with dnf repoquery when possible, otherwise a raw pattern is used.zypper addlock name=version, falling back to addlock name if the version-scoped lock is rejected./etc/apk/world with name=version entries.stages:
default:
- name: "Pin kernel and container runtime"
package_pins:
linux-image-generic: "5.15.0-91.101"
containerd.io: "1.7.11"
stages.<stageID>.[<stepN>].layoutRepartitions a disk: adds partitions in the trailing free space and/or expands the last partition. Sizes are always in MiB; size: 0 means "use all remaining free space".
Device targeting: the target disk is picked by matching (in order) label (filesystem or partition label) or path. path also accepts a script://<command> form — the command is executed and its trimmed stdout is used as the device path, useful when the target is not known ahead of time.
If init_disk: true the disk is wiped and a fresh GPT is written. Destructive — only use on a disk you fully control. disk_name (optional) makes the GPT GUID deterministic (UUIDv5 derived from the name under the URL namespace, uuid.NamespaceURL).
Only the last partition can be expanded, and expansion happens before any new partition is added. Expansion only grows the partition — it never shrinks it.
For add_partitions:
0 means all remaining free space.pLabel already exists it is left untouched, otherwise data may be lost.ext2 (default), ext3, ext4, fat, vfat, fat16, fat32, xfs, btrfs, swap, or noformat / - / none to create the partition without formatting.stages:
default:
- name: "Repart disk"
layout:
device:
label: COS_RECOVERY
path: /dev/sda
init_disk: true
disk_name: "MYDISK"
expand_partition:
size: 4096 # 0 for all remaining free space
add_partitions:
- fsLabel: COS_STATE
size: 8192
pLabel: state
- fsLabel: COS_PERSISTENT
filesystem: ext4
size: 0
- fsLabel: COS_GRUB
size: 512
filesystem: fat32
bootable: true
- pLabel: raw_data
size: 1024
filesystem: noformat
stages.<stageID>.[<stepN>].unpack_imagesUnpacks OCI images onto the filesystem. platform is optional and defaults to the host platform.
stages:
default:
- name: "Unpack images"
unpack_images:
- source: "quay.io/luet/base:latest"
target: "/usr/local/luet/"
- source: "rancher/k3s:latest"
target: "/usr/local/k3s-arm64"
platform: "linux/arm64"
stages.<stageID>.[<stepN>].datasourceFetches user data from the specified cloud providers. The list is iterated in order and the first provider that succeeds is used. Provider-specific data is written under /run/config; user data is stored at path.
stages:
default:
- name: "Fetch cloud user data"
datasource:
providers:
- "aws"
- "digitalocean"
path: "/etc/cloud-data"
stages.<stageID>.[<stepN>].commandsRuns arbitrary shell commands. Executes after file/directory writes so newly written files and directories are available.
stages:
default:
- name: "Regenerate initrd"
commands:
- dracut -f
- depmod -a
mudler/yip is a public repo on GitHub by mudler. :pushpin: Yaml Instructions Processor - Simply applies a cloud-init style yaml file to the system.
It declares the Apache-2.0 license on GitHub.
Its maker has not tagged a system, so it is unknown here. Its README says what it needs.
Open "Take it" on its page: Take My Repo writes an install plan for your machine and hands it to your AI agent (Claude Code, Codex or Cursor), which asks you before each step. Take My Repo has not tested it on any machine.
Its last change on GitHub was on 2026-11-19. Take My Repo does not judge from that whether it is maintained.
Nobody has reviewed it on Take My Repo yet, which says nothing about its quality.

LocalAI is the open-source AI engine. Run any model - LLMs, vision, voice, image, video - on any hardware. No GPU required.

:sailboat: The immutable, decentralized, statically built p2p VPN without any central server and automatic discovery! Create decentralized introspectable tunnels over p2p with shared tokens

LocalAGI is a powerful, self-hostable AI Agent platform designed for maximum privacy and flexibility. A complete drop-in replacement for OpenAI's Responses APIs with advanced agentic capabilities. No clouds. Local AI that works on consumer-grade hardware (CPU and GPU).

:brain: 100% Local Memory layer and Knowledge base for agents with WebUI