Template Library Reference

The MATIH Template Library provides pre-built, parameterized starting points for common platform tasks. Templates accelerate onboarding, enforce best practices, and ensure consistency across tenant workspaces. This section catalogs every template category with detailed descriptions, parameters, and usage instructions.

Overview

Templates are stored in the templates/ directory of the platform repository, organized by discipline:

templates/
  bi/                 # Business Intelligence templates
  data/               # Data Engineering templates
  ml/                 # Machine Learning templates
  agentic/            # AI Agent templates
  ontology/           # Ontology and knowledge graph templates
  notebooks/          # Jupyter Notebook templates
  spark/              # Apache Spark job templates

Templates can be accessed through:

Marketplace UI: Browse and install templates from the Config Service marketplace
CLI: matih template apply <template-name> --params key=value
API: POST /api/v1/marketplace/templates/{id}/apply
Workbench: Template gallery in each workbench application

BI Templates (`templates/bi/`)

Business Intelligence templates provide pre-configured dashboard layouts, widget configurations, and data visualization patterns.

Dashboard Templates

Template	Description	Widgets	Data Requirements
`executive-summary`	C-suite overview dashboard with KPI cards, trend lines, and regional breakdowns	8 widgets	Revenue, cost, user counts, time series data
`sales-pipeline`	Sales funnel visualization with conversion rates and deal stages	6 widgets	CRM data with pipeline stages and amounts
`customer-analytics`	Customer segmentation, cohort analysis, and retention metrics	10 widgets	Customer data with acquisition dates and behavior
`financial-overview`	P&L statement visualization, budget vs actual, cash flow	7 widgets	Financial data with accounts and periods
`product-metrics`	Product usage analytics, feature adoption, and engagement	9 widgets	Product telemetry and user activity data
`operational-kpis`	Operations dashboard with SLAs, throughput, and error rates	8 widgets	Operational metrics and incident data
`marketing-performance`	Campaign performance, channel attribution, and ROI	7 widgets	Marketing campaign data with spend and conversions
`hr-analytics`	Headcount, attrition, compensation, and diversity metrics	8 widgets	HR data with employee demographics and compensation

Template Parameters (BI)

Each BI template accepts the following standard parameters:

Parameter	Type	Required	Description
`name`	string	Yes	Dashboard name
`dataSource`	string	Yes	Connector or data source to bind to
`dateColumn`	string	Yes	Primary date/time column for time series
`dateRange`	string	No	Default date range (`last_7_days`, `last_30_days`, `last_quarter`, `ytd`)
`currency`	string	No	Currency code for financial formatting (default: `USD`)
`refreshInterval`	integer	No	Auto-refresh interval in seconds (default: `300`)
`theme`	string	No	Visual theme (`light`, `dark`, `corporate`)

Usage Example

POST /api/v1/marketplace/templates/executive-summary/apply
Content-Type: application/json
 
{
  "name": "Q4 2025 Executive Summary",
  "params": {
    "dataSource": "sales_warehouse",
    "dateColumn": "order_date",
    "dateRange": "last_quarter",
    "currency": "USD",
    "theme": "corporate",
    "revenueTable": "sales.orders",
    "revenueColumn": "total_amount",
    "regionColumn": "sales_region",
    "productColumn": "product_category"
  }
}

Widget Type Reference

Widget Type	Description	Configuration Keys
`kpi-card`	Single numeric KPI with trend indicator	`metric`, `comparison`, `format`, `thresholds`
`line-chart`	Time series line chart	`xAxis`, `yAxis`, `series`, `groupBy`
`bar-chart`	Vertical or horizontal bar chart	`xAxis`, `yAxis`, `orientation`, `stacked`
`pie-chart`	Pie or donut chart	`dimension`, `measure`, `donut`, `showLabels`
`area-chart`	Stacked or regular area chart	`xAxis`, `yAxis`, `series`, `stacked`
`scatter-plot`	Scatter plot with optional regression line	`xAxis`, `yAxis`, `colorBy`, `sizeBy`
`data-table`	Tabular data with sorting and pagination	`columns`, `sortBy`, `pageSize`, `conditionalFormatting`
`pivot-table`	Pivot table with drill-down	`rows`, `columns`, `values`, `aggregation`
`funnel-chart`	Funnel/conversion chart	`stages`, `measure`, `showConversion`
`map`	Geographic map with data overlay	`geoColumn`, `measure`, `mapType`, `colorScale`
`heatmap`	Two-dimensional heatmap	`xAxis`, `yAxis`, `value`, `colorScale`
`gauge`	Gauge/speedometer chart	`value`, `min`, `max`, `thresholds`
`text-block`	Rich text annotation	`content`, `format`
`filter-control`	Interactive filter widget	`column`, `filterType`, `defaultValue`

Data Engineering Templates (`templates/data/`)

Data engineering templates provide pipeline definitions, transformation patterns, and data quality configurations.

Pipeline Templates

Template	Description	Technology	Schedule
`elt-basic`	Basic ELT pipeline: extract from source, load to staging, transform to target	Temporal + dbt	Hourly
`elt-incremental`	Incremental ELT with change tracking and merge	Temporal + dbt	Every 15 minutes
`cdc-pipeline`	CDC pipeline using Flink to stream database changes to Iceberg	Flink SQL	Continuous
`data-warehouse-load`	Full data warehouse load with dimension/fact table patterns	Temporal + Spark	Daily
`data-lake-ingestion`	Ingest files from S3/Azure Blob/GCS into Iceberg tables	Spark + Airflow	On arrival
`api-to-lake`	Extract data from REST APIs and load into the lakehouse	Python + Temporal	Configurable
`streaming-aggregation`	Real-time aggregation of event streams into materialized views	Flink SQL	Continuous
`data-compaction`	Scheduled Iceberg table compaction and snapshot management	Spark	Daily

Pipeline Template Parameters

Parameter	Type	Required	Description
`name`	string	Yes	Pipeline name
`sourceConnector`	string	Yes	Source data connector name
`targetCatalog`	string	Yes	Target Iceberg catalog
`targetSchema`	string	Yes	Target schema/namespace
`schedule`	string	No	Cron expression for scheduling
`notifyOnFailure`	boolean	No	Send notification on pipeline failure
`retryCount`	integer	No	Number of retry attempts on failure (default: `3`)
`timeout`	integer	No	Pipeline timeout in minutes (default: `60`)

dbt Project Templates

Template	Description	Models
`dbt-starter`	Basic dbt project with staging, intermediate, and mart layers	5 example models
`dbt-ecommerce`	E-commerce analytics dbt project	15 models (customers, orders, products, sessions)
`dbt-saas-metrics`	SaaS metrics dbt project (MRR, churn, LTV, cohorts)	12 models
`dbt-financial`	Financial reporting dbt project (P&L, balance sheet, cash flow)	10 models

Data Quality Templates

Template	Description	Rules
`quality-basic`	Basic data quality checks (nulls, uniqueness, range)	5 rule types
`quality-comprehensive`	Full quality suite with statistical profiling and anomaly detection	15 rule types
`quality-freshness`	Data freshness and SLA monitoring	Freshness checks + SLA definitions
`quality-schema-drift`	Schema drift detection and alerting	Schema comparison rules

ML Templates (`templates/ml/`)

Machine learning templates provide experiment configurations, training scripts, model serving definitions, and monitoring dashboards.

Experiment Templates

Template	Description	Framework	Task
`classification-tabular`	Binary/multiclass classification on tabular data	scikit-learn, XGBoost	Classification
`regression-tabular`	Regression on tabular data	scikit-learn, LightGBM	Regression
`timeseries-forecasting`	Time series forecasting with multiple models	Prophet, ARIMA, LSTM	Forecasting
`nlp-text-classification`	Text classification with transformer models	Hugging Face, PyTorch	NLP
`image-classification`	Image classification with CNN/ViT	PyTorch, torchvision	Computer Vision
`recommendation-engine`	Collaborative filtering recommendation system	PyTorch, Surprise	RecSys
`anomaly-detection`	Unsupervised anomaly detection	scikit-learn, PyOD	Anomaly Detection
`clustering-analysis`	Customer/data segmentation via clustering	scikit-learn, HDBSCAN	Clustering

Training Job Templates

Template	Description	Resources
`training-single-gpu`	Single GPU training job	1 GPU, 4 CPU, 16Gi memory
`training-multi-gpu`	Multi-GPU distributed training	4 GPUs, 16 CPU, 64Gi memory
`training-ray-distributed`	Ray-based distributed training	Ray cluster (1 head + 4 workers)
`training-cpu-only`	CPU-only training for traditional ML	8 CPU, 32Gi memory
`hyperparameter-sweep`	Hyperparameter tuning with Ray Tune	Configurable (2-16 workers)

Model Serving Templates

Template	Description	Serving Technology
`serve-fastapi`	Simple model serving with FastAPI	FastAPI + Uvicorn
`serve-triton`	High-performance serving with NVIDIA Triton	Triton Inference Server
`serve-ray`	Scalable serving with Ray Serve	Ray Serve
`serve-batch`	Batch inference pipeline	Spark + Temporal
`serve-ab-test`	A/B testing deployment with traffic splitting	Ray Serve + Istio

Model Monitoring Templates

Template	Description	Metrics
`monitor-basic`	Basic model performance monitoring	Accuracy, latency, throughput
`monitor-drift`	Data and concept drift detection	PSI, KS test, JS divergence
`monitor-fairness`	Fairness and bias monitoring	Demographic parity, equalized odds
`monitor-comprehensive`	Full monitoring suite	All of the above + custom metrics

ML Template Parameters

Parameter	Type	Required	Description
`experimentName`	string	Yes	MLflow experiment name
`modelName`	string	Yes	Model registry name
`datasetPath`	string	Yes	Path to training dataset (S3/ADLS/GCS)
`targetColumn`	string	Yes	Target variable column name
`featureColumns`	list	No	Feature column names (default: all non-target)
`testSize`	float	No	Test set proportion (default: `0.2`)
`randomSeed`	integer	No	Random seed for reproducibility (default: `42`)
`gpuCount`	integer	No	Number of GPUs (default: `0`)

Agentic Templates (`templates/agentic/`)

Agentic templates provide pre-configured AI agent workflows and conversation patterns.

Agent Templates

Template	Description	Agents
`chat-basic`	Basic conversational Q&A agent	Intent classifier, RAG retriever, response generator
`text-to-sql-standard`	Standard text-to-SQL pipeline	Intent classifier, schema retriever, SQL generator, validator, executor
`text-to-sql-advanced`	Advanced text-to-SQL with disambiguation and visualization	All standard + disambiguator, visualizer, explainer
`data-analyst`	Autonomous data analyst agent	Full pipeline + automated follow-up questions
`report-generator`	Automated report generation from natural language	Full pipeline + report formatter, PDF generator
`custom-domain`	Template for building domain-specific agents	Base agent scaffold with custom tool integration

Agent Template Parameters

Parameter	Type	Required	Description
`name`	string	Yes	Agent workflow name
`defaultDataSource`	string	Yes	Default data source for queries
`defaultDialect`	string	No	SQL dialect (default: `trino`)
`llmProvider`	string	No	LLM provider override (default: tenant setting)
`llmModel`	string	No	LLM model override
`enableGuardrails`	boolean	No	Enable safety guardrails (default: `true`)
`enableStreaming`	boolean	No	Enable response streaming (default: `true`)
`maxRetries`	integer	No	Max SQL generation retries (default: `3`)

Ontology Templates (`templates/ontology/`)

Ontology templates provide domain-specific knowledge graph schemas.

Template	Description	Entities	Relationships
`enterprise-data`	General enterprise data ontology	Database, Schema, Table, Column, User, Application	contains, owns, uses, produces
`ecommerce`	E-commerce domain ontology	Customer, Product, Order, Category, Review, Seller	purchases, contains, reviews, belongs_to
`financial-services`	Financial services ontology	Account, Transaction, Customer, Portfolio, Risk	holds, transfers, owns, manages
`healthcare`	Healthcare domain ontology	Patient, Provider, Encounter, Diagnosis, Medication	treats, prescribes, diagnoses, refers
`saas-product`	SaaS product analytics ontology	User, Account, Feature, Session, Event, Subscription	uses, triggers, subscribes, belongs_to
`iot-sensor`	IoT sensor data ontology	Device, Sensor, Reading, Location, Alert	measures, located_at, triggers, contains

Ontology Template Parameters

Parameter	Type	Required	Description
`name`	string	Yes	Ontology name
`namespace`	string	Yes	Ontology namespace URI
`format`	string	No	Output format (`owl`, `rdf`, `json-ld`, default: `owl`)
`includeShacl`	boolean	No	Include SHACL validation shapes (default: `true`)
`mapToSchema`	string	No	Database schema to auto-map entity-to-table

Notebook Templates (`templates/notebooks/`)

Jupyter Notebook templates provide interactive analysis starting points.

Template	Description	Language	Libraries
`eda-basic`	Basic exploratory data analysis	Python	pandas, matplotlib, seaborn
`eda-advanced`	Advanced EDA with profiling and correlation	Python	pandas-profiling, plotly, scipy
`sql-analysis`	SQL-based analysis with Trino	Python + SQL	trino-python-client, pandas
`ml-experiment`	ML experiment notebook with MLflow tracking	Python	scikit-learn, mlflow, matplotlib
`deep-learning`	Deep learning training notebook	Python	PyTorch, torchvision, tensorboard
`nlp-analysis`	NLP text analysis and visualization	Python	transformers, spacy, wordcloud
`geospatial`	Geospatial data analysis	Python	geopandas, folium, shapely
`timeseries`	Time series analysis and forecasting	Python	statsmodels, prophet, plotly
`spark-analysis`	Distributed analysis with Spark	PySpark	pyspark, spark-connect
`data-quality-report`	Data quality assessment report	Python	great-expectations, pandas

Notebook Template Parameters

Parameter	Type	Required	Description
`name`	string	Yes	Notebook file name
`dataSource`	string	Yes	Data source connection name
`tableName`	string	No	Default table to analyze
`outputFormat`	string	No	Output format for reports (`html`, `pdf`)
`kernelSpec`	string	No	Jupyter kernel (`python3`, `pyspark`)

Spark Job Templates (`templates/spark/`)

Apache Spark job templates provide production-ready Spark application scaffolds.

Template	Description	Language	Mode
`batch-etl`	Batch ETL job: extract, transform, load to Iceberg	Scala/Python	Batch
`streaming-etl`	Structured Streaming ETL from Kafka to Iceberg	Scala/Python	Streaming
`data-compaction`	Iceberg table maintenance (compaction, snapshot expiry)	Scala	Batch
`feature-engineering`	Feature computation and materialization to Feast	Python	Batch
`data-validation`	Large-scale data validation with custom rules	Python	Batch
`graph-analytics`	GraphX-based graph analytics	Scala	Batch
`delta-migration`	Migrate data from Delta Lake to Iceberg	Scala	Batch

Spark Template Parameters

Parameter	Type	Required	Description
`name`	string	Yes	Spark application name
`mainClass`	string	Yes (Scala)	Main class fully qualified name
`mainFile`	string	Yes (Python)	Path to main Python file
`driverCores`	integer	No	Driver CPU cores (default: `2`)
`driverMemory`	string	No	Driver memory (default: `4g`)
`executorCores`	integer	No	Executor CPU cores (default: `4`)
`executorMemory`	string	No	Executor memory (default: `8g`)
`executorInstances`	integer	No	Number of executors (default: `2`)
`inputPath`	string	Yes	Input data path
`outputPath`	string	Yes	Output data path
`icebergCatalog`	string	No	Iceberg catalog name (default: `iceberg`)

Template Versioning

Templates follow semantic versioning. When a template is updated, existing instances are not automatically modified. Users can check for template updates via:

GET /api/v1/marketplace/templates/{id}/versions

Updating an existing instance to a new template version is done via:

POST /api/v1/marketplace/templates/{id}/upgrade
{
  "instanceId": "inst-abc123",
  "targetVersion": "2.0.0",
  "preserveCustomizations": true
}

Custom Templates

Teams can create and publish custom templates to the Marketplace:

Create a template definition in the appropriate templates/ subdirectory
Include a template-manifest.yaml with metadata, parameters schema, and documentation
Submit via POST /api/v1/marketplace/templates
Templates undergo review before publication

# template-manifest.yaml
name: custom-sales-dashboard
displayName: Sales Dashboard (Custom)
version: 1.0.0
category: bi
author: Sales Analytics Team
description: Custom sales dashboard with regional breakdown and YoY comparison
parameters:
  - name: dataSource
    type: string
    required: true
    description: Sales data connector
  - name: revenueColumn
    type: string
    required: true
    default: total_amount
  - name: dateColumn
    type: string
    required: true
    default: order_date
tags: [sales, revenue, dashboard, bi]

Template Application Workflow

When a user applies a template, the platform follows this workflow:

Step-by-Step Process

Step	Action	Component
1	User browses or searches for templates	Marketplace UI or API
2	User selects a template and reviews its description and parameters	Workbench gallery
3	User provides parameter values (data source, names, options)	Parameter form
4	Platform validates parameters against the template's JSON Schema	Config Service
5	Platform renders the template with the provided parameters	Template engine
6	Platform creates the resulting resources (dashboard, pipeline, experiment, etc.)	Target service
7	Platform confirms creation and provides a link to the new resource	Workbench UI

Template Rendering Engine

Templates use a Mustache-compatible rendering engine with the following built-in helpers:

Helper	Description	Example
`{{param}}`	Simple parameter substitution	`{{name}}` renders to `My Dashboard`
`{{#if param}}...{{/if}}`	Conditional rendering	`{{#if enableStreaming}}streaming: true{{/if}}`
`{{#each items}}...{{/each}}`	Iteration over arrays	`{{#each columns}}{{name}}: {{type}}{{/each}}`
`{{uppercase param}}`	Convert to uppercase	`{{uppercase name}}` renders to `MY_DASHBOARD`
`{{lowercase param}}`	Convert to lowercase	`{{lowercase name}}` renders to `my_dashboard`
`{{slugify param}}`	Convert to URL-safe slug	`{{slugify name}}` renders to `my-dashboard`
`{{timestamp}}`	Current ISO 8601 timestamp	`2026-02-12T10:30:00.000Z`
`{{uuid}}`	Generate a new UUID	`a1b2c3d4-e5f6-7890-abcd-ef1234567890`

Template Testing

Before publishing a template, run the template test suite:

cd templates/{category}/{template-name}
matih template test --params test-params.json

The test suite validates:

Check	Description
Schema Validation	All required parameters are defined and typed correctly
Rendering	Template renders without errors with test parameter values
Resource Validation	Rendered resources pass the target service's validation (e.g., valid dashboard JSON, valid pipeline DAG)
Idempotency	Applying the same template twice with the same parameters does not create duplicate resources
Parameter Boundaries	Template handles edge cases (empty strings, maximum values, special characters)

Template Categories Summary

Category	Directory	Count	Target Service
Business Intelligence	`templates/bi/`	8 dashboard templates	BI Service
Data Engineering	`templates/data/`	8 pipeline + 4 dbt + 4 quality templates	Pipeline Service, Data Quality Service
Machine Learning	`templates/ml/`	8 experiment + 5 training + 5 serving + 4 monitoring templates	ML Service
Agentic AI	`templates/agentic/`	6 agent workflow templates	AI Service
Ontology	`templates/ontology/`	10 domain ontology templates	Ontology Service
Notebooks	`templates/notebooks/`	10 Jupyter notebook templates	JupyterHub
Spark Jobs	`templates/spark/`	7 Spark application templates	Spark Operator

Total: 79 templates across 7 categories

Each template is maintained as part of the platform repository and follows the same CI/CD process as application code: automated testing on pull requests, versioned releases, and documentation updates.

Connector SDK Guide Version History