Built for the Xsolla Baku GameTech Hackathon by Team Enthuzone.

In modern game development, studios suffer from two disconnected silos of player feedback:

1. Qualitative Noise: Chaotic player chatter across Discord, Steam reviews, and bug reports that are difficult to categorize, deduplicate, and prioritize.
2. Quantitative Blindness: Massive telemetry logs tracking player deaths and quits without explaining why players are frustrated.

GoBuster bridges this gap. By coupling Google Gemini 2.5 Flash with deterministic mathematical correlation engines, the platform ingests live community feedback from Discord alongside real-time gameplay telemetry from Unity 3D, correlates player complaints with empirical dropoff anomalies, and delivers prioritized, evidence-backed issues and actionable patch advice to Game Designers, Product Managers, and QA Leads.

flowchart TD
    subgraph Sources["🎮 Game Edge & Player Community"]
        Unity["🕹️ Unity 3D Game Client\n(URP Runner + TelemetrySender.cs)"]
        Discord["💬 Discord Community\n(#feedback / #bug-reports)"]
    end

    subgraph IngestLayer["⚡ Forwarding & Relay"]
        Bot["🤖 Discord Bot Relay (:Node.js)\n(Sequential Queue, Exponential Backoff)"]
    end

    subgraph CoreBackend["🖥️ Core Platform Server (:4000)"]
        IngestAPI["📥 Ingestion Gateway\n(/api/platform/ingest/*)"]
        Auth["🔐 Multi-Tenant Auth\n(JWT, RBAC, MongoDB Transactions)"]
        MongoDB[("🍃 MongoDB Database\n(Clusters, Feedback, Telemetry)")]
        ClusterEngine["📊 Issue Clustering & Evidence Linking"]
    end

    subgraph IntelligenceLayer["🧠 AI Intelligence Service (:4001)"]
        Gemini["☁️ Google Gemini 2.5 Flash\n(Structured JSON Output, Injection Defense)"]
        MathEngine["📐 Anomaly & Correlation Engine\n(NFKC Normalization, Roman Numeral Matchers)"]
    end

    subgraph StudioSurface["💻 Operations Dashboard (:3000)"]
        Dashboard["📊 Next.js 16 Web Dashboard\n(Triage Queue, A/B Build Compare, Landing)"]
        StudioStaff["👥 Game Designers, PMs & QA Leads"]
    end

    Unity -- "POST /api/platform/ingest/telemetry\n(x-api-key)" --> IngestAPI
    Discord --> Bot
    Bot -- "POST /api/platform/ingest/discord/:id\n(x-webhook-token)" --> IngestAPI
    
    IngestAPI --> MongoDB
    IngestAPI -- "Async Queue" --> ClusterEngine
    ClusterEngine <--> IntelligenceLayer
    ClusterEngine --> MongoDB

    Dashboard <--> CoreBackend
    StudioStaff <--> Dashboard

All architecture, user flow, API, and setup documentation are organized in the docs/ hub:

Express.js backend with MongoDB replica-set transactions, multi-tenant RBAC, and ingestion webhooks.

cd server
npm install
cp .env.example .env
npm start # Runs on http://localhost:4000

Microservice running Google Gemini 2.5 Flash for structured JSON classification and deterministic correlation math.

cd ai-service
npm install
cp .env.example .env # Set GEMINI_API_KEY
npm start # Runs on http://localhost:4001

Daemon monitoring Discord channels for player feedback and relaying to the Core Server with automatic retries and 🎮 reactions.

cd discord-bot
npm install
cp .env.example .env # Set DISCORD_BOT_TOKEN
npm start

Unity URP action-runner game with automated runtime TelemetrySender.cs streaming player attempts and abandonment metrics.

• Open game in Unity Hub (2022.3 LTS or 2023.2+).
• Open Assets/Scenes/MainScene.unity and press Play.

Next.js 16 web application for triage, release validation, A/B build comparison, and product marketing.

cd client/dashboard
npm install
cp .env.example .env
npm run dev # Runs on http://localhost:3000

1. Player Experience & Feedback: A player reaches Level 5 in the Unity game, experiences excessive boss HP, fails repeatedly, and quits. Frustrated, the player posts in the studio's Discord: "Boss in Level V is impossible to beat, HP pool is absurd!"
2. Real-time Ingestion:
   • Unity's TelemetrySender emits attempt and quit events for level_5 to the Core Server.
   • The Discord Bot picks up the chat message, forwards it via authenticated webhook, and reacts with 🎮.
3. AI Classification & Normalization:
   • The AI Service invokes Gemini 2.5 Flash with strict JSON schema outputs: classifies as Difficulty, assigns severity 0.88, and validates authenticity.
   • Normalization converts "Level V" to canonical target "Level 5".
4. Empirical Correlation:
   • The correlation engine calculates an empirical 42% Abandonment Rate (anomaly $>30%$) and 7.4 Average Retries (anomaly $>5.0$).
   • The issue is tagged as Supported by Telemetry with a 94% Confidence Score, escalating it to CRITICAL.
5. Studio Decision & Patch:
   • The QA Lead views the paired evidence on /issues.
   • Clicks AI Recommendation: Gemini suggests: "Reduce Boss Phase 2 HP by 15-20% and add an intermediate checkpoint."
   • The studio deploys Build 1.1.0 and uses /compare to verify a 65% drop in complaints and a 24% increase in level completion.

Built with passion for the Xsolla Baku GameTech Hackathon (Sept 9–11, 2026). For questions or inquiries, please refer to the team leads or open an issue in this repository.