AI Script Installer

Installing a FiveM resource by hand means reading the README, running the SQL, editing server.cfg, registering jobs and items in your framework, and chasing down dependencies. The AI Script Installer reads the resource, plans that work as a set of reviewable actions, and shows you exactly what it will change before it changes anything.

How it works

Add a resource, from the marketplace or by dropping a folder or archive onto the installer, and Boltise moves through five states: uploading, analyzing, preview, installing, done. The AI analysis streams in while it reads the resource and detects your framework (ESX, QBCore, QBox, VORP, RSG, or standalone).

When analysis finishes you land on the review screen, a two-column preview:

• Left, the plan. A verdict, what the script does, the Setup Actions Boltise will run, plus integration notes and recommendations.
• Right, the evidence. The resource’s dependencies pinned at the top, then tabs for Files, SQL, Config, and Security.

Nothing is written to disk or your database until you choose to apply.

The verdict

Every analysis ends with one of three verdicts, shown at the top of the plan:

Verdict	Meaning
Ready to Install	No blockers. Apply whenever you want.
Review Before Installing	Works, but read the warnings first (missing dependency, destructive SQL).
Cannot Install	A blocker (for example a failed security scan). Resolve it before installing.

Setup actions

A setup action is one reviewable change. Boltise plans them as typed actions, each with a title and a short rationale:

Type	What it does
job-register	Adds a job and its grades to your framework’s jobs table (shared/jobs.lua).
item-register	Adds items to your framework’s items table (shared/items.lua).
config-edit	Sets values in a resource config (for example config.lua).
sql-run	Runs the resource’s SQL against your connected database.
manual	A step Boltise cannot automate, with a link or instructions.

Actions carry flags: required ones block the install until they are handled, and destructive ones (any sql-run) ask for confirmation before they touch your database.

Choose how much Boltise does for you with the mode switch:

• Plan preview and apply each action yourself.
• Run All run everything in order, pausing on destructive steps.
• Auto run the safe actions, skip the destructive ones.

Each row moves through Reviewed, Running…, then Applied, Skipped, or Failed, and the header keeps a running count like 2/5 done · 3 auto. If the resource needs nothing, you will see Everything's already configured and can install straight away.

Example: an ambulance job on QBCore

For a typical ambulance job, Boltise detects QBCore and plans three actions:

• job-register the ambulance job with its grades, written into qb-core/shared/jobs.lua.
• item-register the bandage and medikit items, written into qb-core/shared/items.lua.
• sql-run the resource’s own .sql file, which creates the tables it needs.

The first two are file edits, not database writes — Boltise finds the jobs or items table in your framework, splices in a Lua block, and validates that the file still parses before saving it. Only sql-run touches your database.

Note

job-register and item-register need a Lua table to write into (QBShared.Jobs, QBX.Shared.Jobs, Config.Jobs, or a bare Jobs). On a framework that keeps jobs in the database rather than in a Lua file, the AI plans the equivalent work as sql-run or manual instead.

Files, SQL, Config, Security

The right column is the evidence behind the plan. Dependencies stay pinned at the top, with anything missing flagged (the ambulance job needs es_extended and oxmysql). Below that:

• Files the resource files that will be copied into resources/.
• SQL a summary of the database work: how many tables will be created and modified, each new table with its columns, and the total number of rows to be inserted.
• Config the config values Boltise will set.
• Security the result of the static scan that can move the verdict to Cannot Install.

The SQL tab summarises what sql-run will do, rather than printing the file:

create	2
modify	1
inserts	14

New tables are listed with their columns, modified tables by name. Anything destructive — a DROP or a TRUNCATE in the resource’s SQL — raises an amber warning banner above the counts and moves the verdict to Review Before Installing.

And the server.cfg change is shown as a diff before it is written:

server.cfg

ensure es_extended
ensure oxmysql
ensure esx_ambulancejob

Applying, backups, and revert

When you install, Boltise runs a fixed pipeline and reports each step:

1. Backup existing resource snapshot anything it is about to overwrite.
2. Copy resource files write the resource into resources/.
3. Execute SQL files run the approved SQL.
4. Update server.cfg add the ensure line.
5. Finalizing confirm the result.

Every file Boltise writes is snapshotted for the session, so a successful action shows a Revert that restores the previous file.

Caution

SQL is the one step a file backup cannot undo. Before a destructive action runs, Boltise warns: this will modify your database and cannot be auto-reverted, it snapshots the file but database changes are permanent. Read the SQL tab first, and keep your own database backups on production servers.

Supported frameworks

The installer understands the conventions of ESX, QBCore, QBox, VORP, and RSG, and falls back to Standalone for resources with no framework dependency (it still handles their files, SQL, and server.cfg). The detected framework appears in the review header alongside the file count and size, for example ESX · 42 files · 1.2 MB.

Tips

• Keep BOLTISE.md current. The planner reads your project rules, so noting your framework and conventions sharpens the plan.
• Check the Dependencies and Security panels before installing. A missing dependency or a failed scan is what moves a verdict off Ready to Install.
• Use Plan mode on a production server so you apply each change deliberately. Run All is for resources you already trust.

Related

• Installing a FiveM Resource — check requirements, SQL, startup and the actual in-game behavior.

• AI Code Review — review your own changes with AI backed by static analysis before you push.

• Code Editor — edit installed resources with FiveM/RedM native autocomplete.