🌌 TDS Exam Portal — Elite Workspace

🌟 The Vision

TDS Portal is a production-grade, highly secure, browser-based sandbox environment engineered for the automated execution, verification, and forensic analysis of deterministic computational solvers. Built specifically as an advanced study companion for the IIT Madras Tools in Data Science (TDS) degree curriculum, the portal bridges the gap between intricate, multi-layered data science pipelines and reproducible, instantaneous client-side computations.

Constructed around a core philosophy of visual excellence, academic transparency, and mathematical determinism, the platform registers and orchestrates 162+ specialized solvers across 14 exam suites, delivering sub-millisecond output vectors, interactive diagnostic notes, and automated rubric evaluations 100% client-side.

🏗️ Core Architectural Pillars

The workspace relies on four highly optimized client-side subsystems to maintain absolute security, speed, and mathematical rigor:

graph TD
    A["User Interface (Glassmorphism Dark IDE)"] -->|"Normalized Input (Email + Token)"| B("app.js (Core Orchestrator)")
    B -->|"Active Checkbox Guard"| C{"Academic Integrity Guard"}
    C -->|"Pass (Read Storage)"| D["Dynamic ESM Import"]
    C -->|"Fail (Block Action)"| A
    D -->|"Import Registry"| E["Term Registry Map (T1/T2)"]
    E -->|"Load Target Solver"| F["Shared runtime.js"]
    F -->|"Access Lock Check"| G{"lockConfig.locked?"}
    G -->|"Authorized / Open"| H("Seeded Variations Engine (FNV-1a / PRNG)")
    G -->|"Restricted / External"| I["Lock Screen + Rubric Coach Terminal"]
    H -->|"Deterministic Solved Vector"| J(("Deterministic Case Solution"))
    J -->|"Native DOM Mount"| K["Rendered Answer + Notes Panel"]
    I -->|"Native DOM Mount"| L["Interactive Rubric Coach Evaluator"]
    K --> A
    L --> A
style A fill:#121214,stroke:#374151,stroke-width:2px,color:#fff
    style C fill:#451a03,stroke:#ea580c,stroke-width:2px,color:#fbbf24
    style G fill:#2e1065,stroke:#8b5cf6,stroke-width:2px,color:#c084fc
    style H fill:#064e3b,stroke:#059669,stroke-width:2px,color:#34d399
    style J fill:#1e1b4b,stroke:#4f46e5,stroke-width:2px,color:#818cf8
    style L fill:#831843,stroke:#db2777,stroke-width:2px,color:#f472b6

1. Seeded Mathematical Invariance

• normalizeEmail(): Standardizes user-supplied emails by trimming trailing whitespace and casting all characters to lowercase.
• rng() / createRng(): Cryptographically stable, seeded pseudo-random number generator wrapped around seedrandom.js and fast FNV-1a 32-bit hashing. Using the standardized email as the entropy seed guarantees that every shuffle, parameter selection, variation synthesis, and index split is 100% identical for a given student across different browsers or systems.

2. Combinatorial Variations Engine

• Plagiarism Prevention: Permutes sentence formulations, evidentiary row orderings, and hypothesis refutations uniquely per student seed (${email}:${questionId}).
• Fact Invariance: Guarantees that despite lexical divergence, 100% of underlying ground-truth data points (batch IDs, file paths, monetary sums, timestamp anomalies, and confidence ratings) remain intact.

3. Multi-Term ESM Dynamic Registry

• Decoupled Modules: Solver files are encapsulated in modular ES6 classes that only load when an exam is selected.
• Structured Interface Contract: Every resolver must adhere to a strict structural interface:

interface SolverResponse {
      id?: string;                 // Canonical question identifier
      title: string;               // Display title
      answer: string;              // The literal computed solution / raw text
      type: "solved" | "guide" | "bypass" | "error"; 
      variant: string;             // Variant identification metadata
      answerDisplay: string;       // Rich formatted Markdown preview
      guide?: string;              // Implementation walkthroughs for CLI steps
      rubricCoachHtml?: string;    // Interactive Rubric Coach terminal component
      debug?: Record<string, any>; // Computational trace logs & benchmark stats
    }

4. Interactive Rubric Intelligence Terminal

• Live Heuristic Parsing: Analyzes user-drafted Markdown in real-time, validating heading skeletons (## Judgment, ## Evidence Table, etc.), character count gates ([150–3,000] or [200–6,000]), and evidence matrix columns (| Claim | Source | Confidence |).
• Citation Density Matrix: Scans draft text for high-impact entities (e.g. DF-00020, Farah Iqbal, INV-17, ~17.5%, 9021.10.00, SUP-02, 02:10:00, $123,565.26).
• Live Scorecard: Generates an instant rubric score (0–100%) with categorized passed checks, unmet criteria, and strategic recommendations.

🔒 Academic Integrity & Access Security

To align fully with the IIT Madras Student Code of Conduct, the workspace implements strict multi-layer protections:

[!WARNING]

MALPRACTICE ALERT: This sandbox is strictly designed for local educational research, study, algorithm verification, and rubric calibration. Using this portal during active, live, proctored, or graded exams is a direct violation of university rules and constitutes academic dishonesty.

1. Mandatory Integrity Disclaimer: All workspace inputs and solver executions are blocked until the user explicitly acknowledges the honor code checkbox.
2. Configurable Access Locking (lock-config.js): Granular per-exam security locks allow developers to restrict programmatic answer generation to authorized whitelist emails while providing external students with the interactive Rubric Coach to verify their own manual work.
3. Stateful Local Persistence: The acceptance vector is saved to browser localStorage as academic_integrity_agreed === 'true'.

🎯 Complete Exam Registry & Computational Engines

The registry includes Term 1 2026 (T12026), Term 2 2026 (T22026), and Term 3 2026 (T32026; official term dates TBD). T3 GA1 follows the official September 2026 question bundle and preserves all 16 question slots. Five deterministic questions have local answer generation; three API deployment questions are explicit handoffs to the separate API project; other live-personal tasks are shown as guides rather than guessed answers. T1/T2 solver suites and the dedicated 300-question T2 end-term practice workspace are available:

| Target Engine | Term | Scope | Technical Highlights & Capabilities | | :--- | :--- | :--- | :--- | | GA1 | T32026 | Developer Tools (September 2026) | 16 official question slots with a pinned source snapshot. Seeded generators for Q3/Q4/Q5/Q7/Q8/Q11; live API URLs for Q6/Q14 and an on-demand Q15 form (private Ledger link → live URL, filled into the Answer box); a local Q10 Git-history finder (upload/paste history.fi); Q12 live-tested prompts plus an in-browser tester (your own aipipe.org token, never stored); guides for Q2/Q9/Q13/Q16 where the student must act. | | P1 | T32026 | Project 1 (October–November 2026) | 4 interactive questions, each with a tool that takes the student's own input and a beginner step-by-step guide: an AI Pipe usage forecast with exact LMSR trade maths, a 60-second Human Exchange script builder, an Air Ink phone-readiness check (no AI agents), and an exact invoice calculator with a request screen. Nothing uses the private launch links. | | End-Term Mock | T22026 | May 2026 Full-Course Practice | 300-question pool: 175 MCQs, 25 MSQs, and 100 subjective prompts, presented in a dedicated quiz UI with section jumping, answer checking, a five-step subjective guide, detailed explanations, model answers, self-assessment rubrics, bookmarks, timer, and per-email local progress. Reference-only practice material; no score guarantee. | | P2 | T22026 | Project 2 — Forensic Case Studies | 8 Case Studies (100 Marks): DTH Month-End Mystery (DF-00020/$0.00), DTH Complaints Quiet (NovaIVR/CareDesk dropouts, 5 diagnostic questions), Solar Inverter Smell Test (INV-17/0 MW telemetry), Solar 31.6% DSM Claim (within-day ~17.5% counterfactual decomposition), Swiss Customs Mismatch (tariff 9021.10.00 schema mapping), QC Queue Smell Test (02:10:00 AM QCore snapshot / 48h LabTrack latency), Irish Preference Claim (SUP-02 / P1006 / IE-2025-000411 2025 LTSD), and Spare-Parts Search ($123,565.26 inventory risk classification). Includes the full Rubric Intelligence Terminal. | | GA8 | T22026 | LLM Eval, Guardrails & MLOps | 10 Solvers: Automated RAG hallucination scanners, LLM benchmark pipelines, semantic guardrail firewalls, model drift analyzers, and Docker/Cloud Run deployment scripts. | | GA7 | T22026 | Policy Gates & OSINT | 10 Solvers: 5 hosted rule-engine APIs (release gates, LLM action firewall, Terraform policy, output sanitizer, OSINT corroboration); 4 seeded generators (Google-dork queries, WAF rule-order simulation, media forensics, GitHub Actions audit); Street View image intelligence. | | GA6 | T22026 | Data Forensics & Automation | 10 Solvers: Seeded prompt-robustness audit, shadow-DOM incident audit, DuckDB ledger reconciliation, politeness/robots.txt audit, client-side Canvas 2D tile reconstruction, Web Audio FFT burst decoding, and interactive SQL generators. | | GA5 | T22026 | Agentic Systems & Guardrails | 11 Solvers: MCP servers, A2A invoice protocols, budget guards, skill scanning, redteam/guardrail APIs, incident-response agents, mailroom triage, proration, and personal backup endpoints for Q9/Q10/Q11. | | GA4 | T22026 | RAG & Retrieval Engineering | 13 Solvers: Hybrid-search chunking, RRF fusion, HyDE, GraphRAG pipelines, late chunking, ANN recall/latency tuning, semantic caching, multimodal embedding calibration, and vector-rerank APIs. | | GA3 | T22026 | System & API Architecture | 13 Solvers: Automated video metadata curation, multimodal QA, fixed/dynamic schema extraction, cosine similarity mapping, browser-based Proof-of-Work nonce miner, Context Window Heist extractor, and asciinema pipelines. | | GA2 | T22026 | API & Cloud | 10 Solvers: Descriptive statistics with CORS, OAuth JWT verification, 12-factor config precedence, multi-container Redis stacks, POST analytics aggregation, Prometheus observability, and rate-limiting. | | GA1 | T22026 | Developer Tools | 20 Solvers: Git history analysis, VS Code multi-cursor edits, shell pipelines, SQL schemas, Markdown layout docs, HTTP POST requests, and asciinema recordings. | | GA0 | T22026 | Intro to Data Science | 25 Solvers: Standard exam solvers, detailed ngrok setups, FastAPI templates, CORS configurations, and forensic sandboxing guides. | | P1 | T22026 | Project 1 | 4 Solvers: Requirements-interview guide, seeded model-intelligence differentiation solver, and Colab AI-agent guides with shared service-account pools. | | ROE | T22026 | Re-Exam Suite | 12 Solvers: Incident-atlas routing, Unicode ledger forensics, HTTP cache simulation with standalone offline Node.js CLI fallbacks, byte-verified hash/HMAC transcriptions, and audio/essay pre-flight checkers. | | ROE | T12026 | Re-exam Workflows | 15 Solvers: Procedural maze pathfinding, regex golf parsing, and programmatic arithmetic validation. | | GA7 / GA8 / P2| T12026 | Data Viz, MLOps & Solana | 15 Solvers: Midpoint diverging palette sampler, chartjunk analyzers, Docker verification hashes, Cloud Run triggers, and Solana devnet QR transaction repair + Discourse facts indexer. |

🏠 Home Page Solver Index, SEO & Dynamic Config

The home page lists every term, exam and question the portal supports, so students and search engines can see the whole project without running a solver.

• Source of truth: tds-config.json (terms, their exams, the term label such as "T3 2026 (Sep–Dec)") plus each exam's registry.js. Nothing about terms or years is hard-coded in the generator.
• Generator: node scripts/build-catalog.mjs rewrites marker blocks in index.html (CATALOG, SEO, TERMS), the intro description (visible text, meta, Open Graph, Twitter and tds-config.json), and writes catalog.json, sitemap.xml and robots.txt. --check fails if anything is stale (line endings ignored); check.mjs runs it.
• Latest additions: the newest work of the newest term (up to four cards), filled from the previous term when it has fewer than four. Term status (Current / Recent / Archive) and colours follow recency, not term ids.
• Dynamic UI pieces: the T1 · T2 · T3 pills under the title and the sidebar Term selector come from the config; new terms appear without edits. Exams missing from the label table in the script get a label derived from their id.
• Deploy: vercel.json runs the generator on every build (buildCommand, falling back to the committed files if it fails), so a push is enough to refresh the index, sitemap and descriptions.
• Still manual for a new exam: add it to tds-config.json, the exam menu (TERM_EXAMS in app.js) and create solvers///registry.js.
• Wording rule: generated copy must not contain the words "answer(s)" or "solution(s)" (IITM academic-integrity expectations); the generator throws if it does.
• Creator card: photo (hover popup with GitHub and LinkedIn links), live GitHub follower and star counts (cached for an hour, hidden if the API is unreachable), and LinkedIn/GitHub/repo chips.
• Production: https://tds-portal-solver.vercel.app (Vercel; main deploys to production, dev gets a preview).

🎨 Premium Glassmorphic UI & Workspace Experience

The user interface is crafted to mirror an elite dark-themed IDE workspace:

• Glassmorphic Terminal Panels: High-performance CSS glassmorphism leveraging backdrop-filter: blur(16px), reactive focus glows matching the active color theme, and sub-pixel typography.
• Answer-First Visual Flow: Solved answers are displayed immediately in high-contrast monospaced blocks, while implementation guides collapse cleanly below.
• Instant Utilities: One-click Copy Answer, Copy Variant, Download as .txt, and Focus Mode toggles.
• Rubric Draft Toolkit: Fast clipboard ingestion (📋 Paste), template loading (📑 Insert Skeleton), draft resetting (🧹 Clear), and live character/word counters with reactive boundary indicators (Green for valid gates, Red for violations).
• Vibe Audio Studio: Built-in, non-intrusive ambient audio player with customizable local track loading via IndexedDB, custom equalizers, and theme-matched glowing animations.

📂 Codebase Architecture Blueprint

tds-roe-solver/
├── 📂 .agents/                   # Agent behavioral profiles and configurations
├── 📂 solvers/
│   ├── 📂 T12026/                # Term 1 2026 Exam Registry (GA7, GA8, ROE, P2)
│   │   ├── 📂 ga7/               # Midpoint color generators & standard RNG hooks
│   │   ├── 📂 ga8/               # Docker verification & Cloud Run triggers
│   │   ├── 📂 roe/               # Maze pathfinding & regex golf solvers
│   │   └── 📂 p2/                # Solana devnet QR & Discourse facts indexer
│   ├── 📂 T22026/                # Term 2 2026 Exam Registry (GA0–GA8 + ROE + P1 + P2)
│   │   ├── 📂 ga0/ ... 📂 ga8/   # Modular deterministic solvers (GA0 through GA8)
│   │   ├── 📂 p1/                # Project 1 AI Agent & GCS setups
│   │   ├── 📂 p2/                # Project 2 Case Study Engine & Rubric Coach
│   │   │   ├── 📄 case-specs.js  # Case boundaries, headings & rubric criteria
│   │   │   ├── 📄 rubric-coach.js# Rubric Coach terminal & draft evaluator
│   │   │   ├── 📄 lock-config.js # Access control configuration & whitelist
│   │   │   ├── 📄 runtime.js     # Solver execution wrapper & diagnostics
│   │   │   ├── 📄 variations-engine.js # PRNG variation & table generator
│   │   │   ├── 📄 registry.js    # Ordered question registry (1-to-1 live exam match)
│   │   │   ├── 📄 q-case-1a.js   # Case 1A: DTH Month-End Mystery
│   │   │   ├── 📄 q-case-1b.js   # Case 1B: DTH Complaints Went Quiet
│   │   │   ├── 📄 q-case-2a.js   # Case 2A: Solar Inverter Smell Test
│   │   │   ├── 📄 q-case-2b.js   # Case 2B: Solar 31.6% Impact Claim
│   │   │   ├── 📄 q-case-3a.js   # Case 3A: Swiss Mismatch Control
│   │   │   ├── 📄 q-case-4a.js   # Case 4A: QC Queue Smell Test
│   │   │   ├── 📄 q-case-3b.js   # Case 3B: Irish Preference Claim
│   │   │   └── 📄 q-case-4b.js   # Case 4B: Spare-Parts Search
│   └── 📂 T32026/                # Term 3 (GA1 Developer Tools & GA2 API Engineering)
│       ├── 📂 ga1/               # GA1 16 official slots (Tangram SVG, Git forensics, Ledger)
│       └── 📂 ga2/               # GA2 15 official slots (CORS, OAuth, Observability, Ollama)
├── 📂 scratch/                   # Automated Monte Carlo stress test harnesses
├── 📂 scripts/
│   └── 📄 build-catalog.mjs      # Generates the home-page index, SEO files and term selector
├── 📄 app.js                     # Central UI state machine & event orchestrator
├── 📄 check.mjs                  # Comprehensive multi-term test suite (14 modules)
├── 📄 index.html                 # Main portal shell (contains generated CATALOG/SEO/TERMS blocks)
├── 📄 catalog.json               # Generated machine-readable solver index
├── 📄 sitemap.xml / robots.txt   # Generated SEO files
├── 📄 tds-config.json            # Terms, exams and intro text (source of truth for the home page)
├── 📄 vercel.json                # Build command (runs the generator) and cache headers
├── 📄 server.js                  # Traversal-safe lightweight local static server
├── 📄 style.css                  # Master stylesheets & Rubric Coach terminal styles
├── 📄 sw.js                      # Progressive Web App offline Service Worker
└── 📄 verify.html                # Universal Solver Verification Hub Dashboard

🚀 Quickstart & Developer Hub

1. Spinning Up the Portal

npm install
npm run dev

2. Running the Full Test Suite

npm run check

3. Running the Project 2 Rubric & Offline Evaluation Stress Test

node scratch/stress-test-raw-generator.js

4. Generating the T2 2026 End-Term Mock Bank

npm run mock:check
npm run mock:generate

Reference-only notice: this mock is original practice material, not an official IIT Madras paper, answer key, or prediction. It cannot guarantee a high end-term score and is intended only for revision and last-minute practice.

Crafted with ❤️ for the IITM TDS Research & Learning Community.