# AI Configuration Generator

Source: https://docs.ops0.com/docs/configurations/ai-generator

ops0's Configurations feature turns plain-English requests into production-ready Ansible playbooks, roles, and inventory — generated by Claude and scoped to the specific servers or server group you're targeting.

![ops0 AI Configuration Generator showing generated Ansible files ready for review](https://docs.ops0.com/images/ai-generator-chat.png)

---

## Creating an Ansible Configuration Project

### Open Configurations
Navigate to **Configurations** and click to create a new project.

### Choose Project Type
Select **Ansible** as the project type (the alternative is **Kubernetes**, for Helm/manifest projects).

### Choose Target Infrastructure
Pick a target mode: **Server Group** (target every server in a group) or **Individual Servers** (hand-pick specific servers from your [Servers](https://docs.ops0.com/docs/servers/overview) fleet).

### Name and Create
Give the project a name and description, then create it. This opens the project's conversational editor.

---

## Generating a Playbook

Inside a project, you describe what you want in natural language and the AI responds with an explanation and the generated files.

### Describe what you need
Type a request describing the configuration you want, for example: "Install and configure nginx as a reverse proxy for port 3000" or "Harden SSH: disable password auth and root login."

### The AI detects your intent
ops0 first classifies your message as either a **question** (asking about existing configuration) or a **change** (asking to create or modify files). Questions get a direct explanation with no files touched; changes proceed to generation.

### Files are generated and streamed in
For a change request, the AI streams back the playbook and any supporting files — roles, inventory, group/host variables, Jinja2 templates, and handlers — as they're produced.

### Review the explanation and confirm
The AI returns a short explanation of what will change. Changes require confirmation before they're applied to the project.

**Targeted edits, not just full rewrites**

For small changes to a large existing file, the generator can return a targeted patch instead of rewriting the whole file, so a one-line change to an existing playbook doesn't regenerate everything around it.

---

## What Informs the Generated Playbook

| Input | How it's used |
|-------|----------------|
| **Your natural-language request** | The primary instruction for what to build or change |
| **Existing project files** | Included as context so edits are consistent with what's already there |
| **Target servers or server group** | Server names, IP addresses, and detected operating systems are passed in so generated tasks match your real infrastructure |
| **Enabled policies** | Any policies enabled on the project are enforced as mandatory constraints — the AI is instructed to generate compliant code or warn you if your request would violate one |

---

## Ansible Best Practices Applied

The generator is instructed to consistently follow:

- **Structure** — a `site.yml` entry point with roles organized under `roles/`
- **Idempotency** — tasks that are safe to re-run
- **Security** — no hardcoded secrets; use of `ansible-vault` where secrets are needed
- **Documentation** — meaningful, descriptive task names
- **File organization** — playbooks, roles, inventory, `group_vars`/`host_vars`, Jinja2 templates, and handlers are all generated as distinct, correctly-typed files

---

## Generated File Types

| Type | Example location |
|------|-------------------|
| **Playbook** | `site.yml` and other root-level `*.yml` playbooks |
| **Role** | `roles/<name>/tasks/*.yml` |
| **Inventory** | `inventory/*.yml` |
| **Vars** | `group_vars/*.yml`, `host_vars/*.yml` |
| **Template** | `roles/<name>/templates/*.j2` |
| **Handler** | `roles/<name>/handlers/*.yml` |

---

## Questions vs. Changes

You can also use the same chat to ask about your configuration without changing anything — for example, "Is my playbook idempotent?" or "What does this role do?" The AI answers directly with no files modified, which makes it safe to sanity-check a playbook before deploying it.

---

## Next Steps

- [Servers Overview](https://docs.ops0.com/docs/servers/overview): Add the servers your playbooks will target

- [Server Groups](https://docs.ops0.com/docs/servers/groups): Target a whole group of servers at once