Modliq Documentation
Practical, beginner-friendly guides for using Modliq. Learn how to upload data, connect databases, write goals, run optimization, validate quality, track operations, and generate buyer-ready Quality Passports.
Start here: the 6-step Modliq workflow
From raw data ingestion to buyer-ready quality documentation, follow this step-by-step path.
Create a Project
Create a project for a plant, product line, process, or customer use case. Each project keeps its own datasets, goals, results, Quality Studio analysis, and Quality Passport.
Upload or Connect Data
Upload CSV or Excel files, extract tables from PDF/Word documents, or connect read-only databases such as Supabase/Postgres and MongoDB.
Review Dataset Health
Modliq checks missing values, outliers, duplicates, identifier columns, target leakage risks, and sample size before optimization.
Write a Goal
Describe what you want in plain English, such as “maximize yield while keeping temperature below 90°C.”
Review and Confirm
Before optimization, confirm the detected target, controllable variables, constraints, and safety acknowledgement.
Run, Validate, and Report
Run optimization, validate stability in Quality Studio, review operations and supplier risks, and generate a Quality Passport.
Using Modliq for Education and Research
Modliq can be used to teach and learn EDA, visualization, dataset health, feature importance, model comparison, and report generation without requiring a full Python setup.
First EDA Demo Guide
Learn how to upload sample CSV datasets, run automated exploratory data analysis, generate correlation matrices, and inspect missing value distributions in 2 minutes.
Comparing ML Models without Code
Train Linear Regression, Random Forest, and XGBoost models simultaneously. Evaluate model performance on leaderboard rankings without writing pipeline code.
Understanding R², RMSE, and MAE
Interpret regression metrics (R², Root Mean Squared Error, Mean Absolute Error) and classification metrics (Accuracy, Precision, Recall) in plain language.
Interpreting Feature Importance
Read SHAP feature importance bar charts to explain which input variables drive target predictions in your dataset.
Exporting Project Reports for Research
Export clean, formatted Markdown and PDF research summaries for paper submissions, lab reports, and thesis appendix documentation.
Using Modliq in a Classroom
Step-by-step instructions for educators to set up student cohorts, assign prebuilt classroom lab templates, and demonstrate applied data science without software installation.
How Modliq combines analytics and machine learning
Modliq first helps users understand what happened in their data, similar to the work of a data analyst. It checks dataset quality, summarizes columns, detects outliers, visualizes relationships, and produces quality and operations insights. Then Modliq helps users decide what to try next, similar to the work of an ML engineer. It parses goals, trains models, evaluates performance, recommends process settings, and generates safe trial ranges.
| Business Need | Traditional Role | Modliq Feature |
|---|---|---|
| Clean and understand data | Data Analyst | Dataset Health + EDA Studio |
| Visualize trends & distributions | Data Visualization Eng | Chart Studio (No-Code Charts) |
| Find patterns & correlations | Data Analyst | Correlations, distributions, trends |
| Analyze process capability | Data Analyst / Quality Eng | Quality Studio (SPC, I-MR, Cpk) |
| Analyze plant operations | Data Analyst / Ops Mgr | OEE and Downtime Pareto |
| Define prediction goal | ML Engineer | Natural Language Goal Parser |
| Prepare model features | ML Engineer | Review & Confirm Setup Wizard |
| Train predictive model | ML Engineer | AutoML Optimization Engine |
| Recommend process setpoints | ML Engineer | Constrained Safe Trial Optimization |
| Validate trial output | Process Engineer / Quality | Quality Studio + Trial SOP Generator |
| Document evidence for buyers | Quality Head / Management | Buyer-Ready Quality Passport |
Choosing the right chart for manufacturing data
Chart Studio automatically recommends visual chart types tailored to your dataset structure. Use this quick reference matrix to select the best visual representation for your factory operational question.
| Operational Question | Recommended Chart | Primary Insight |
|---|---|---|
| How has yield changed over time? | Line Chart | Detects shift-by-shift trends, process drift, and temporal stability. |
| Which supplier has the lowest yield? | Bar Chart | Compares categorical averages to highlight vendor variance. |
| What causes most downtime? | Pareto Chart | Ranks failure drivers using the 80/20 rule to prioritize Kaizen actions. |
| Is temperature related to yield? | Scatter Plot | Examines non-linear relationships, sweet spots, and process boundaries. |
| Which variables are correlated? | Correlation Heatmap | Multi-variable Pearson matrix to identify co-linear parameters. |
| Is the process stable? | I-MR Control Chart | Monitors 3-sigma Upper and Lower Control Limit compliance. |
| How do 5S categories compare? | Radar Chart | Multi-axis comparison for Sort, Set in order, Shine, Standardize, Sustain. |
| What is current OEE? | KPI Card | High-impact summary metric for executive decision-making. |
Writing good goals
A good goal tells Modliq what metric to improve and what process limits must be respected. Good goals are specific, measurable, and connected to columns in your dataset.
Good Goal
Why it works:
- Target is clear: yield
- Direction is clear: maximize
- Constraints are clear: temperature and pressure limits
- Columns match dataset: temperature, pressure, yield
Bad Goal
Why it is weak:
- No target metric: unquantified intent
- No direction: missing maximize/minimize instruction
- No process constraints: no operating boundary checks
- No measurable success condition: cannot validate with ML
Goal Examples by Industry Domain
Maximize yield while keeping temperature below 90°C and pressure below 5 bar.Minimize defect rate while keeping moisture between 8% and 12%.Maximize assay result while keeping pH between 6.5 and 7.2.Maximize fermentation yield while keeping pH between 6.8 and 7.2 and dissolved oxygen above 30%.Minimize rejection rate while maintaining dimension within specification limits.Column name examples and what Modliq detects
Modliq reads your column names to identify targets, process variables, identifiers, traceability fields, quality metrics, and operations fields.
| Column Name Example | Detected As | Notes |
|---|---|---|
| yield | Target / Numeric Metric | Often used as optimization target |
| yield_rate | Target / Numeric Metric | Common yield column |
| temperature | Process Feature | Used as controllable input |
| pressure | Process Feature | Used as controllable input |
| flow_rate | Process Feature | Used as controllable input |
| pH | Process Feature | Used as controllable input |
| batch_id | Identifier | Treated as metadata, not controllable feature |
| lot_id | Identifier / Traceability | Used for traceability, not optimization control |
| supplier | Supply Chain Field | Used for supplier scorecard and risk analysis |
| supplier_lot | Traceability Field | Links material lots to batch outcomes |
| operator_name | Categorical Field | Useful for grouping/filtering |
| timestamp | Datetime Field | Useful for time-series analysis |
| defect_count | Quality Metric | Used for quality and defect analysis |
| reject_count | Quality Metric | Used for rejection and OEE quality rate |
| good_count | Operations Metric | Used for OEE quality rate |
| downtime_minutes | Operations Metric | Used for downtime Pareto |
| downtime_reason | Operations Category | Used for downtime Pareto |
| shift | Operations Category | Used for shift comparison |
| machine | Operations Category | Used for bottleneck analysis |
| scrap_rate | Lean / Quality Metric | Used for scrap and waste insights |
Documentation Topics
Select a module below to jump directly to detailed instructions and recommendations.
Getting Started
Project creation & workflow overview
Education & Research
Classroom EDA, model metrics & research reports
Uploading Data
CSV, Excel, & PDF table ingestion
Connecting Databases
Supabase, Postgres, & MongoDB
Dataset Health
Profiling & readiness scoring
Writing Good Goals
Natural language goal formatting
Running Optimization
AutoML & safe parameter windows
Quality Studio
SPC, I-MR charts, & Cp/Cpk metrics
Operations
OEE, downtime Pareto, & shifts
Supply Chain
Supplier scorecards & lot traceability
Lean
Waste tracking, 5S, & Kaizen boards
Quality Passport
Buyer-ready evidence reporting
AI Copilot
SOP drafts, CAPA, & insights
Security & Privacy
Data protection & credential security
FAQ
Common questions & answers
Getting started with Modliq
Anchor: #getting-started
Start by creating a project. A project represents one manufacturing use case, such as improving yield for a product line, analyzing defects for a batch process, or preparing a Quality Passport for a buyer.
Quick Start Steps
- Sign in to Modliq console.
- Create a new project.
- Load the demo dataset or upload your own factory data.
- Review the automated dataset health score.
- Define your optimization goal in plain English.
- Run optimization and validate results in Quality Studio.
Uploading data
Anchor: #uploading-data
Modliq supports multiple ways to bring manufacturing data into the platform for automated analysis and machine learning optimization.
Supported Data Sources
- CSV files (.csv)
- Excel spreadsheets (.xlsx, .xls)
- PDF or Word documents with tables
- Supabase/Postgres database tables
- MongoDB collections
- Demo manufacturing datasets
File Format Guides
Best Practices for Data Upload
- Use clear column names
- Avoid merged cells in Excel
- Keep one row per batch, lot, or event
- Include target metrics (yield, defects)
- Include process variables (temperature, pressure, flow rate, pH, humidity, cycle time)
Connecting databases
Anchor: #connecting-databases
Modliq supports read-only database connectors for engineering teams that store live production logs in databases.
Supabase / Postgres
Read-only connection via SSL string or parameters.
MongoDB
Read-only collection snapshot ingestion.
MySQL & SQL Server
Enterprise connectors in development.
Connector Safety & Guardrails:
- Connectors operate strictly in read-only mode (SELECT operations only).
- Live preview is capped at 100 rows; snapshot import is capped at 10,000 rows.
- Credentials are encrypted server-side before storage.
- Raw database passwords are never returned to the browser client.
Dataset health check
Anchor: #dataset-health
Before running machine learning optimization, Modliq checks whether your dataset is reliable enough for valid statistical inference.
Automated Health Checks
- Missing values percentage
- Duplicate row count
- Sample size sufficiency
- Outliers via IQR and Z-score
- Zero-variance constant columns
- Suspicious identifier columns
- High collinearity detection
- Target leakage risk warnings
- Automatic target column recommendation
Readiness Score Bands
Note: A high health score means the dataset is structurally suitable for initial model training. It does not guarantee physical production performance without engineering trial validation.
Running optimization
Anchor: #running-optimization
Optimization uses historical production data to recommend process settings that may improve a target metric while respecting physical plant constraints.
Optimization Execution Flow
- Type your natural language goal.
- Modliq parses target, direction, features, and constraints.
- Review and confirm the setup wizard.
- Acknowledge controlled trial safety notice.
- Start optimization execution.
- Monitor training progress.
- Review results and safe trial ranges.
Optimization Engine Outputs
- Recommended process settings
- Safe parameter trial ranges
- SHAP feature driver rankings
- Model accuracy metrics (R², MAE, RMSE)
- Expected yield improvement delta
- Projected financial ROI estimate
- Actual vs predicted scatter plot
- Interactive optimization curve
Quality Studio guide
Anchor: #quality-studio
Quality Studio helps quality teams analyze process stability, statistical process control (SPC), capability indices, and inspection readiness.
Key Features in Quality Studio
Operations guide
Anchor: #operations
Operations tools help production managers track Overall Equipment Effectiveness (OEE), analyze downtime causes, eliminate line bottlenecks, and compare shift performances.
Operations Features
- OEE calculator (Availability, Performance, Quality)
- Downtime Pareto chart
- Production line comparison
- Shift performance comparison
- Machine bottleneck detection
Calculated deterministically from logged operating minutes, theoretical cycle times, and scrap counts.
Supply Chain guide
Anchor: #supply-chain
Supply Chain tools help connect raw material vendor quality and batch lot numbers to plant yield outcomes.
Supply Chain Capabilities
- Supplier scorecard rating
- Material lot end-to-end traceability
- Yield breakdown by vendor
- Defect rate breakdown by material lot
- Incoming raw material quality risk score
- Vendor anomaly alerts
Lean guide
Anchor: #lean
Lean tools help manufacturing teams convert statistical findings into continuous improvement (Kaizen) actions.
Lean Tools
- 8 Waste tracker
- Kaizen action board
- 5S workplace audit scorer
- Takt time calculator
- Kanban sizing calculator
Tracked Waste Categories
Quality Passport guide
Anchor: #quality-passport
The Quality Passport is a buyer-ready evidence report summarizing dataset readiness, process stability, capability metrics, and continuous improvement steps.
What standard Quality Passports include
- Dataset health & sample size
- Optimization summary & safe ranges
- SPC control chart stability
- Cp / Cpk capability indices
- OEE & scrap summary
- Supplier material lot traceability
- Kaizen & CAPA status
- Audit readiness score (0-100)
- Missing evidence warning list
- Recommended SOP next actions
AI Copilot guide
Anchor: #ai-copilot
The AI Copilot helps explain optimization results, draft SOPs, summarize Quality Passports, suggest CAPA root-cause actions, and assist plant managers.
AI Copilot Use Cases
- Goal wording improvement suggestions
- Dataset health warning explanations
- SHAP process driver explanations
- SPC out-of-control rule interpretation
- CAPA action plan drafting
- Standard Operating Procedure (SOP) drafts
- Operations & downtime reviews
- Supplier quality risk explanations
- Quality Passport executive summaries
Deterministic Guardrail Architecture
AI assists. Engineers approve. Deterministic calculations such as OEE, Cp/Cpk, SPC limits, and dataset health scores are computed directly by Modliq's Python ML engine—never invented or hallucinated by AI models.
Security and data privacy
Anchor: #security
Modliq is engineered with zero-trust principles. The frontend browser application communicates exclusively with the Express backend gateway.
Security Principles
- JWT Authentication & Session tokens
- Project-level data access control
- Encrypted database credentials
- Service-key protected ML engine
- Backend-only AI provider keys
- Read-only database connectors
- Private user console isolation
- No public indexing of console pages
Frequently Asked Questions
Find quick answers about Modliq data ingestion, optimization, and platform security.
Ready to explore no-code machine learning?
Whether you are a manufacturer, teacher, student, professor, or research scholar, Modliq helps you explore data and machine learning without code.