PUSH & LIFT

PLIA

The Push & Lift AI: it talks to you, has specialist modes and runs the Suite apps.

What it is for

PLIA is a conversation with an AI that knows iRacing and the Push & Lift apps. It explains the car and the rules, helps with setups, opens and runs MECANIC, GARAJE, SPOTTER, DIRECTOR, TOOLKIT and LIVERY STUDIO, and works with the services you connect.

It needs a brain, meaning an AI model: one from your account, one with an API key, or a local one on your graphics card.

How to open it

  1. In the Suite menu, PIT group, click PLIA.
  2. Click OPEN. It opens in its own window.
  3. The first time takes a little longer because it prepares your profile.

Choosing the brain

You choose it in PLIA's Settings, Brain block. The app may show these labels in your language.

  • CLI: uses your account with an AI assistant you already have. Click Sign in and follow the steps.
  • API: paste your key into the box on the provider's row. Under each key there is a link to where you get it. Models or Edit turn each provider's models on and off. Add provider adds one that is not on the list.
  • Local models: they run on your graphics card with no internet. They are grouped by graphics memory. Click Download and use and it ends up selected. PLIA 1 Mini and PLIA 1 Max are our own, and they are experimental. Once installed you see Use in chat and Remove.

Then, in the message box, Select model changes the model and the effort.

The main screen

Sidebar

  • New session, Search sessions and View options.
  • Add workspace: picks the folder PLIA works in.
  • Hovering a session: Rename, Duplicate session and Archive session.
  • Appearance: colours, backgrounds, styles and effects.
  • Settings.

Header

The session title, the mode, More actions, New tab, Split and Full screen. The Chat and Trajectory tabs switch between the conversation and what PLIA has done.

Message box

  • Choose workspace and Agent preset, which is the mode.
  • Attach, or paste or drop a file.
  • Commands: compact, export, plan, model and others.
  • Access mode: read only, write in the workspace, or full access.
  • Talk, Read messages aloud, Mute microphone and Voice options.
  • Select model and Context usage.
  • Send message and Stop generating.

Note what you test

  1. Tell it what you changed and how it went: "wing at 5 at Spa, better".
  2. PLIA notes it in your notebook as better, same or worse, and reads it when you ask.

On each answer

Copy, Good answer, Bad answer and Branch into a new conversation.

The modes

  • PLIA: the general one. Start here.
  • On track: Setups, Coach, Spotter and My PC.
  • Create: Liveries, Videos and Live.
  • Build: Coder and Factory, for making programs. Factory is in beta.

Settings

  • Brain: CLI, API and Local models.
  • Connections: Paddock Connect, for the simracing services you connect, and Your channel, for your streams and videos. Each card has the link and the steps to get your key. You can also ask it to add your own connector.
  • General: permissions, language, font size, conversation display and what Enter does while it works.
  • Agent presets: the modes, to choose the default, duplicate them or create one.
  • Advanced: the rest, such as plugins and the configuration file.

Common tasks

Asking about iRacing or an app

  1. Write your question in your language.
  2. PLIA searches its library and these manuals, and tells you where to find it, with the link.
  3. If it does not find it, it says so.

Asking for a tutorial

  1. Write, for example: "give me a tutorial on FUEL".
  2. PLIA reads the manual and guides you step by step.

Using an app from the conversation

  1. Ask for it: "open DIRECTOR" or "which cars does MECANIC have".
  2. If needed, it tells you what it is turning on to do it.

Working in another folder

  1. Click Add workspace.
  2. Pick the folder. What you do in that session stays there.

Publishing on your channel

  1. Connect your channel in Connections, Your channel.
  2. Use the Videos or Live mode. Videos upload as private by default.

In depth

Local models by graphics memory

There are groups for 2, 4, 6, 8, 10-16 and 24 GB. With an NVIDIA card, it marks the ones that fit. You can paste another model's name in the Local models box; only download from known authors, because the model drives PLIA's tools on your PC. If it cannot use tools, PLIA rejects it. Small models answer fast but fail more on multi-step tasks.

Topics PLIA turns on

To avoid wasting tokens, PLIA does not carry all its tools. When needed it tells you in one line and turns on the topic: setups, telemetry, apps, accounts, channel, PC, team and spending.

Access modes

  • Read only: it looks and changes nothing.
  • Write in the workspace: it changes files in that folder.
  • Full access: it can do what you ask on your PC, and changes to your accounts or your channel happen right away. Merging a Factory result and marking a video as seen still ask for approval on screen.

Connections

Reading comes first: nothing changes unless you ask. Connected only shows when the service has really answered. The model does not see your keys. They are stored encrypted for your Windows user. To get one, open its card, follow the link and paste the key. With Add connector, PLIA makes you one for another service.

Videos

  1. Choose Videos mode and tell it where the raw footage is.
  2. PLIA removes silences and repeated phrases, and shows you the result.
  3. Watch it with sound and approve it.
  4. On your order it prepares the package: video, thumbnail and sheet. They upload as private.

Factory

  1. Choose Factory mode and say what you want, in your own words.
  2. PLIA turns it into a clear brief and waits for your yes.
  3. It does the work with tests and review.
  4. Merging the result asks for your approval on screen.

Voice

Talk lets you speak to it, and reading messages aloud makes it answer by voice. In the voice options you choose microphone, output, language and voice.

Common problems

  • It says there is no key or credential: that model has no key. Add it in API or choose another.
  • Limit or quota reached: many free models have a cap per minute or per day. Wait or change model.
  • It answers badly or makes things up: it is usually the model. Try a bigger one; small local ones fail more.
  • It stops and waits: check whether it is waiting for your approval on screen. It depends on the Access mode.
  • A local download stops: it is cut if it makes no progress. Click again and it continues where it was.
  • It takes a few seconds to reopen: normal if you have just closed it.
  • Keys are missing after changing PC or reinstalling Windows: they are encrypted for your user and do not travel. Paste them again in API and Connections. They also do not show if you go back to an earlier version; they return when you update.