StateMate
An AI assistant for GTA 5 RP servers that reads rules, analyzes video evidence, and groups answers into clear blocks.
The Beginning
StateMate started as a way to understand API tokens, AI reasoning, and turning a regular text response into specific UI blocks. Instead of an abstract demo, I chose a very narrow scenario: an RP administrator for a small GTA 5 RP server who can answer based on the rules, help with RP texts, and draft forum complaints.
The Context the AI Had to Keep
The project was about a beautiful wrapper around the model, ensuring the AI could work with the server context: rules, legislative RP database, mechanics, forum forms, video evidence, and what is actually visible in the chat and on the HUD.
- The rulebook and RP laws were read as a structured source, not just as a single long text
- The response was broken down into reasoning process, relevant rules, links, and next actions
- The AI had to understand the mechanics of the specific server, rather than answering with general RP phrases
Who StateMate was created for
- Players and admins who need to quickly evaluate a situation according to the rules
- Players who want to formulate RP text, appeals, or forum complaints
- Moderators working with video clips, HUD, chat, and metadata
- Server owners who need a clear knowledge base for supported servers
Rule Engine
Convenient reading of the rulebook, legislative RP database, and server mechanics with links to the required forum sections.
Evidence Input
Inputs of various formats: text, situations, forum templates, and video clips with frame-by-frame analysis.
AI Reasoning
The response was shown not as a single wall of text, but as reasoning, conclusion, rules, next action buttons, and ready drafts.
Account Layer
Firebase for accounts, avatars, names, tokens, request expenses, and a landing page with the status of supported servers.
Where AI Became the Interface
The most interesting part was teaching the model not just to answer, but to assemble the result into components: rule blocks, forum transitions, widgets, next action buttons, and complaint generation. Separately, I experimented with video: the AI analyzed frames, extracted recording time from HUD and metadata, found players, read the chat, and verified everything against server mechanics.
- 01
Connected the model and learned to work with tokens, context, and request costs
- 02
Structured the knowledge base on rules, RP laws, and server mechanics
- 03
Made a dialogue that returns not just text, but a set of widgets and actions
- 04
Added video clip analysis: frames, HUD, chat, metadata, and player behavior
- 05
Built the Firebase layer for accounts, profile, token balance, and supported servers
Successful Test
StateMate had a full product shell: users could change their name and avatar, buy tokens, spend them on different types of requests, and see which servers were supported and how complete their knowledge base was on the landing page. A lot of time was spent on micro-animations so that generation, page transitions, button presses, and dialogues felt alive.