Simulator Guide

This directory contains browser-based learning simulators for practical data, analytics, AI, and technical skills. Each simulator is designed as an interactive practice environment, not a static lesson page. Users can change inputs, run scenarios, inspect outputs, and save local notes or runs where supported.

Shared Behavior

All simulators follow the same dark lab-style UI pattern with scenario cards, metrics, working panels, output/result areas, and local notes or saved runs where useful.

Most simulator data is built into the client component as small practice datasets. These datasets are intentionally lightweight so the simulator is fast, safe, and easy to understand.

Unless a simulator explicitly says otherwise, user work is saved only in the browser using localStorage. It is not saved to the backend database, not attached to a user account, and not shared between devices or browsers. Clearing browser storage will remove saved notes, snippets, and run history.

SQL Practice Simulator

What It Is For

The SQL simulator helps users practice querying relational business datasets. It focuses on the everyday SQL patterns used in analytics, reporting, and data validation.

What Users Can Perform

• Select a built-in dataset such as sales warehouse or support operations.
• Inspect available tables and relationships.
• Run SQL-like SELECT queries against local sample data.
• Practice joins, filters, grouping, ordering, limits, and aggregate calculations.
• Compare result tables with the selected scenario goal.

What Users Can Learn

• How tables connect through keys.
• How to join fact and dimension-style tables.
• How filtering changes query results.
• How grouped aggregations produce business metrics.
• How SQL result sets map to reporting and dashboard questions.

How It Works

The simulator uses local JavaScript data structures and a limited SQL parser/executor inside the browser. It does not connect to a real database and does not run server-side SQL.

Where Data Is Saved

Custom SQL simulator dataset work is stored in browser localStorage under the simulator's local storage key. It remains on the user's device only.

BI Simulator

What It Is For

The BI simulator helps users practice business intelligence thinking: choosing data sources, interpreting KPIs, identifying data-quality risks, and selecting appropriate dashboard visuals.

What Users Can Perform

• Choose business scenarios such as retail, SaaS, or healthcare operations.
• Review KPI cards, trend indicators, data sources, and quality issues.
• Work through BI decision steps.
• Inspect visual summaries such as bars, trend lines, donut charts, and KPI snapshots.
• Produce a short business-facing interpretation of the scenario.

What Users Can Learn

• How BI dashboards connect metrics to decisions.
• How to evaluate source quality and data ownership.
• How KPI targets and trends affect prioritization.
• How to choose visuals for different business questions.
• How to communicate dashboard findings clearly.

How It Works

The simulator uses predefined business scenarios, KPI values, source metadata, and chart data inside the client component. Interactions update the UI locally.

Where Data Is Saved

Any saved BI scenario state is stored in browser localStorage. No BI simulator data is sent to a backend service.

Python Simulator

What It Is For

The Python simulator lets users practice Python in the browser without backend code execution. It is useful for learning Python basics, data manipulation, NumPy, and Pandas examples.

What Users Can Perform

• Load built-in Python examples.
• Edit code in the browser.
• Run Python code.
• View standard output, standard error, return values, and exceptions.
• Stop a running execution by terminating and recreating the worker.
• Save custom snippets locally.

What Users Can Learn

• Core Python syntax and control flow.
• Lists and dictionaries.
• Basic numeric analysis with NumPy.
• DataFrame filtering, grouping, and sorting with Pandas.
• How Python errors and tracebacks appear during execution.

How It Works

Python runs in the browser through Pyodide loaded from a CDN inside a Web Worker. The worker preloads numpy and pandas, captures output, and returns results to the UI. There is no API route or backend Python execution.

The first load requires internet access because Pyodide is loaded from the CDN.

Where Data Is Saved

Current code and saved snippets are stored in browser localStorage. Python code and results are not saved to the server.

Machine Learning Simulator

What It Is For

The ML simulator helps users practice model decision-making rather than model training code. It focuses on business scenarios where threshold tuning and evaluation metrics matter.

What Users Can Perform

• Choose scenarios such as churn prediction, fraud screening, candidate shortlisting, and equipment failure forecasting.
• Switch between model types such as logistic regression, decision tree, and random forest.
• Adjust the classification threshold.
• Inspect predictions, probabilities, actual labels, and outcomes.
• Review precision, recall, accuracy, F1 score, flagged count, and confusion matrix.
• Save experiment runs locally.

What Users Can Learn

• How probability thresholds affect business actions.
• How precision and recall trade off against each other.
• How false positives and false negatives create different costs.
• How to read a confusion matrix.
• How feature importance can support model interpretation.

How It Works

The simulator uses small predefined datasets and simple client-side scoring formulas to mimic classification behavior. It does not train real ML models and does not call an external ML service.

Where Data Is Saved

Experiment notes and saved runs are stored in browser localStorage.

Data Analytics Simulator

What It Is For

The Data Analytics simulator helps users practice the full analyst workflow: data cleaning, KPI calculation, segmentation, chart reading, and recommendation writing.

What Users Can Perform

• Choose datasets such as retail sales, support operations, and marketing funnel data.
• Toggle cleaning steps such as removing duplicate IDs, filling missing values, and keeping only completed work.
• Review data-quality issue counts.
• Group data by region, channel, category, or status.
• Inspect segment charts and clean dataset tables.
• Write analyst notes and save analyses locally.

What Users Can Learn

• How messy records affect KPIs.
• Why duplicate, missing, and incomplete records matter.
• How cleaning choices change metrics.
• How segment-level comparisons support decisions.
• How to explain findings with caveats.

How It Works

The simulator uses built-in tabular datasets and performs cleaning, grouping, KPI calculation, and chart updates directly in the client component.

Where Data Is Saved

Cleaning choices, notes, and saved analyses are stored in browser localStorage.

RAG / Knowledge Base Simulator

What It Is For

The RAG simulator helps users understand retrieval-augmented generation workflows. It focuses on retrieving the right knowledge base context before trusting an answer.

What Users Can Perform

• Choose a knowledge base such as company policy, product support, or learning program documentation.
• Select or edit a user question.
• Adjust chunk size, overlap, top-k, and minimum similarity score.
• Inspect retrieved chunks, matched terms, and source documents.
• Preview a grounded answer based on retrieved context.
• Review grounding, coverage, and risk scores.
• Save retrieval runs locally.

What Users Can Learn

• How chunking affects retrieval.
• Why overlap can improve or dilute context.
• How top-k changes answer coverage.
• How strict score thresholds can hide useful context.
• How retrieved sources affect answer quality and hallucination risk.

How It Works

The simulator uses small local document corpora. It chunks documents in the browser, tokenizes the query, scores chunks with simple lexical matching, and drafts an answer from retrieved context. It does not call an LLM or vector database.

Where Data Is Saved

Queries, retrieval settings, notes, and saved retrieval runs are stored in browser localStorage.

Important Notes

• These simulators are learning tools, not production analytics, AI, or database systems.
• Built-in data is sample data and should not be treated as real customer, employee, financial, or operational data.
• Browser storage is local to the device and browser profile.
• No simulator currently requires authentication to store its local work.
• Python execution is client-only through Pyodide and depends on CDN availability.
• RAG answer previews are simulated from retrieved text and do not use a live LLM.
• SQL execution is intentionally limited to supported practice patterns.

Suggested User Flow

1. Choose a simulator based on the skill being practiced.
2. Select a scenario or dataset.
3. Change the available controls.
4. Inspect the output, metrics, charts, or retrieved context.
5. Write notes explaining the decision.
6. Save the run or analysis locally if the simulator supports it.
7. Refresh or revisit the page to continue from browser-saved state.