e20439b544
All checks were successful
Smoke Test / smoke (pull_request) Successful in 7s
- Adds ansible/ deploy_turboquant.yml playbook with per-node config - Adds turboquant-deploy role: OS-specific (darwin/debian) tasks - Adds health_check.sh and integration test (chat completion) - Adds inventory.ini.example with Mac/Allegro/Ezra groups - Deploys llama.cpp with TurboQuant (Metal on macOS) - Systemd service (Linux) with TURBO_LAYER_ADAPTIVE env
93 lines
2.2 KiB
YAML
93 lines
2.2 KiB
YAML
---
|
|
# Debian/Ubuntu deployment — installs llama.cpp with TurboQuant, uses systemd
|
|
|
|
- name: Update apt cache
|
|
apt:
|
|
update_cache: yes
|
|
cache_valid_time: 3600
|
|
tags: [turboquant, deps]
|
|
|
|
- name: Install build dependencies
|
|
apt:
|
|
name:
|
|
- build-essential
|
|
- cmake
|
|
- git
|
|
- curl
|
|
- python3
|
|
- python3-pip
|
|
- python3-venv
|
|
state: present
|
|
tags: [turboquant, deps]
|
|
|
|
- name: Create turboquant user
|
|
user:
|
|
name: "{{ turboquant_user }}"
|
|
system: yes
|
|
shell: /usr/sbin/nologin
|
|
create_home: no
|
|
tags: [turboquant, prereq]
|
|
|
|
- name: Create install directory
|
|
file:
|
|
path: "{{ turboquant_install_dir }}"
|
|
state: directory
|
|
mode: '0755'
|
|
owner: "{{ turboquant_user }}"
|
|
group: "{{ turboquant_user }}"
|
|
tags: [turboquant, prereq]
|
|
|
|
- name: Clone llama.cpp TurboQuant fork
|
|
git:
|
|
repo: "https://github.com/TheTom/llama-cpp-turboquant.git"
|
|
dest: "{{ turboquant_install_dir }}/llama.cpp"
|
|
version: "feature/turboquant-kv-cache"
|
|
force: yes
|
|
tags: [turboquant, source]
|
|
|
|
- name: Build llama.cpp with TurboQuant
|
|
shell: |
|
|
cd {{ turboquant_install_dir }}/llama.cpp
|
|
cmake -B build -DCMAKE_BUILD_TYPE=Release
|
|
cmake --build build -j$(nproc)
|
|
args:
|
|
creates: "{{ turboquant_install_dir }}/llama.cpp/build/bin/llama-server"
|
|
tags: [turboquant, build]
|
|
|
|
- name: Create models directory
|
|
file:
|
|
path: "{{ turboquant_install_dir }}/models"
|
|
state: directory
|
|
mode: '0755'
|
|
owner: "{{ turboquant_user }}"
|
|
group: "{{ turboquant_user }}"
|
|
tags: [turboquant, deploy]
|
|
|
|
- name: Deploy systemd service unit
|
|
template:
|
|
src: turboquant.service.j2
|
|
dest: /etc/systemd/system/{{ turboquant_service_name }}.service
|
|
mode: '0644'
|
|
tags: [turboquant, service]
|
|
|
|
- name: Reload systemd daemon
|
|
systemd:
|
|
daemon_reload: yes
|
|
tags: [turboquant, service]
|
|
|
|
- name: Enable and start TurboQuant service
|
|
systemd:
|
|
name: "{{ turboquant_service_name }}"
|
|
state: started
|
|
enabled: yes
|
|
tags: [turboquant, service]
|
|
|
|
- name: Deploy health check script
|
|
copy:
|
|
src: "../../health_check.sh"
|
|
dest: "{{ turboquant_install_dir }}/health_check.sh"
|
|
mode: '0755'
|
|
owner: "{{ turboquant_user }}"
|
|
group: "{{ turboquant_user }}"
|
|
tags: [turboquant, deploy]
|