Skip to content

Ticket Masala: System Capabilities & Architectural Overview

Ticket Masala is a next-generation Modular Monolith designed to bridge the gap between enterprise ERP systems (like SAP) and agile, AI-augmented operations. This document provides a detailed synthesis of the system's core capabilities, innovations, and operational principles.

1. Core Architectural Pillars

Modular Monolith First

Ticket Masala follows a "Monolith First" approach for simplicity and performance, while maintaining strict logical boundaries between components. - Single Container Deployment: Minimal DevOps overhead. - SQLite Performance Doctrine: Leveraging Write-Ahead Logging (WAL) and FTS5 for high-speed local data operations without external database dependencies. - In-Process Integration: Low-latency communication between core services and the AI engine.

Multi-Tenant & Multi-Domain

The system supports a two-tier configuration model: - Tenants (Organization Level): Complete data isolation and branding for different companies or departments. - Domains (Process Level): Unique workflows (IT, HR, Landscaping) sharing the same tenant infrastructure but with distinct rules, terminologies, and AI strategies.

2. The Configuration Engine (DSL Compiler)

Innovation: "Compile, Don't Interpret"

Instead of slow runtime logic checks, Ticket Masala uses a Domain-Specific Language (DSL) based on YAML that is compiled into C# Expression Trees at startup.

  • Dynamic Terminology: Change "Ticket" to "Incident," "Permit," or "Planting Request" via config.
  • Stateless Rules: Business logic is defined as high-performance delegates (Func<Ticket, bool>), ensuring <1ms execution time.
  • Hot-Reload: Configurations can be updated and reloaded without restarting the application.
  • Versioning: Every configuration snapshot is hashed (SHA256) and tracked, ensuring that historical decisions are auditable even after rules change.

3. GERDA AI Dispatch Engine

The **G**rouping, **E**valuation, **R**anking, and **D**ispatch **A**lgorithm (GERDA) is the intelligence hub of the system.

Key AI Components:

  • WSJF (Weighted Shortest Job First): Prioritizes work based on business value, time criticality, and risk reduction divided by effort.
  • Affinity Routing: Automatically matches repeat customers with the same agent to improve continuity and satisfaction.
  • Skill-Based Matching: Uses proficiency-level requirements to ensure the right agent handles the right complexity.
  • Workload Balancing: Actively prevents burnout by penalizing assignments to agents over 80% utilization.
  • Explainable AI: Every dispatch suggestion includes a detailed breakdown (e.g., "+50 Affinity", "-20 Workload") so team leads understand the "why."

4. Scalable Ingestion & SAP Integration

Gatekeeper API (High-Throughput)

A dedicated Minimal API project designed for Event-Driven Intake. - Asynchronous Pipeline: Uses System.Threading.Channels to accept 100k+ webhooks/second without blocking. - Scriban Templating: Powerful mapping engine to transform raw external JSON/CSV into the internal domain model. - Background Processing: Decouples the "Acceptance" of data from its "Processing" and "Storage."

SAP Snapshot Sync (On-Demand)

A "Read-Only Amplifier" strategy that eliminates "Excel Hell" while keeping the ERP as the single source of truth. - Immutable Snapshots: Creates versioned states of SAP data linked to specific work items. - "Time Travel" Audits: The ability to see exactly what the data looked like at the moment a dispatch decision was made.

5. Privacy & Governance Proxy

Innovation: "The Compliance Fortress"

Ticket Masala enables safe adoption of LLMs (like OpenAI/Azure) by localizing privacy and cost controls.

  • Local PII Scrubber: Automatically detects and redacts sensitive data (NISS, VAT, IBAN, Email) locally before it ever reaches a cloud API.
  • Ephemeral AI Pipeline: Processes documents (OCR → Summarize → Suggest) in memory; extracted binary blobs are discarded immediately to keep the database lean.
  • Budget Governance: Hard and soft caps on API spending per user and per tenant to prevent "bill shock."
  • Audit Trail: Comprehensive logs of every AI interaction, stored in a GDPR-compliant, scrubbed format.

6. Twitter-Style Knowledge Base

Innovation: Atomic Self-Ranking Streams

A lightweight replacement for traditional stale wikis, focused on friction-free contribution.

  • Atomic Snippets: Knowledge units are the size of a tweet (50-300 words).
  • #Hashtag Organization: No complex folder hierarchies; just tag and search.
  • MasalaRank Algorithm: Content is ranked by: Usage Count + (Expert Verification × 5) - Age Decay
  • Implicit Feedback: Snippets that successfully help close tickets automatically rise to the top of search results.
  • AI Context Injection: The enrichment pipeline automatically pulls relevant KB snippets into LLM prompts for grounded, domain-specific AI suggestions.

7. Dashboards & Insights

  • Team Lead Atelier: A command center for reviewing AI dispatch recommendations, overriding decisions, and monitoring real-time capacity.
  • Manager Dashboard: High-level trends on SLA compliance, agent performance, and domain-specific throughput.
  • Compliance Dashboard: Real-time visibility into AI costs, redacted PII counts, and governance status.

This document synthesizes features and guides as of December 2025.