Features Overview

Actyze delivers enterprise-grade features for secure, real-time analytics across your live data — without ETL pipelines, vendor lock-in, or feature gates. Every feature listed below is free and unlimited under the AGPL v3 open-source license.

Core Capabilities

Natural Language to SQL

Ask questions in plain English (or 50+ other languages) and get accurate SQL queries. Recommended: Claude Sonnet 4.5, GPT-4o, or equivalent high-performance models for best accuracy.

"Show me top 10 customers by revenue in Q4"
↓
SELECT customer_name, SUM(revenue) as total_revenue
FROM sales
WHERE quarter = 'Q4'
GROUP BY customer_name
ORDER BY total_revenue DESC
LIMIT 10

Multilingual Support

Query your data in 50+ languages using our advanced multilingual semantic search:

Supported Languages:

• European: English, German, French, Spanish, Italian, Portuguese, Dutch, Polish, Russian, Czech, Swedish, Danish, Norwegian, Finnish, Romanian, Bulgarian, Greek, Ukrainian
• Asian: Chinese (Simplified & Traditional), Japanese, Korean, Thai, Vietnamese, Indonesian, Malay, Hindi, Bengali, Tamil, Telugu, Marathi, Urdu, Nepali
• Middle Eastern: Arabic, Hebrew, Persian (Farsi), Turkish
• Others: Afrikaans, Albanian, Azerbaijani, Basque, Belarusian, Bosnian, Catalan, Croatian, Estonian, Galician, Georgian, Gujarati, Hungarian, Icelandic, Irish, Kannada, Kazakh, Kurdish, Kyrgyz, Latvian, Lithuanian, Macedonian, Malayalam, Maltese, Marathi, Mongolian, Punjabi, Serbian, Sinhala, Slovak, Slovenian, Somali, Swahili, Tagalog, Tajik, Tatar, Uzbek, Welsh, Yiddish

Powered by:

• Multilingual MPNet (paraphrase-multilingual-mpnet-base-v2) - Primary semantic search engine that understands queries across all 50+ languages
• spaCy NER (English) - Lightweight entity extraction for enhanced query understanding

The heavy lifting is performed by our multilingual embedding model, enabling accurate schema matching regardless of the query language.

Intelligent Schema Detection

FAISS-powered semantic search combined with spaCy Named Entity Recognition automatically identifies the right tables and columns for your queries.

Multi-Database Support

Connect to multiple data sources through Trino connectors:

• Relational: PostgreSQL, MySQL, Oracle, SQL Server
• Cloud Warehouses: Snowflake, Databricks, BigQuery, Redshift
• NoSQL: MongoDB, Cassandra, Elasticsearch
• Data Lakes: Iceberg, Delta Lake, Hudi

→ Database Connectors Guide

Key Features

Scheduled KPIs (Gold Layer)

Define SQL-based KPIs that are collected on a schedule (every 1-24 hours) and materialized as real typed Postgres tables in the kpi_data schema. Each KPI creates a first-class queryable table discoverable by the AI via FAISS indexing.

• Define any SQL query as a recurring KPI
• Collection interval from 1 to 24 hours
• Results materialized as real typed columns (not JSONB blobs)
• Auto-registered with FAISS for AI/natural language discovery
• Time-series data with collected_at timestamp for trend analysis
• Import SQL from saved queries
• RBAC enforced: ADMIN and USER can create, READONLY can view

Use cases: pre-aggregated daily revenue, hourly active users, periodic inventory snapshots — like Databricks gold tables or Power BI import mode, using PostgreSQL.

Predictive Intelligence

Create ML-powered predictions from your KPI data — no data science knowledge required. Choose what you want to predict, select your data, and the system handles model selection, training, and output automatically.

How it works:

1. Choose a prediction type: Forecast (future values), Classify (group membership), Estimate (expected numbers), or Detect (anomalies)
2. Select data source: an existing KPI or a custom SQL query (supports multi-table JOINs via Trino)
3. System auto-validates data quality, selects the best ML model, and trains
4. Predictions stored in prediction_data schema — queryable via NL→SQL just like any other table

Models (per-image, independently deployable via Helm):

• XGBoost — tabular classification, regression, and anomaly detection via Isolation Forest (churn, fraud, CLV, scoring, outlier identification)
• LightGBM — fast tabular predictions, preferred for large datasets (>100K rows)
• AutoGluon TimeSeries — multivariate ensemble forecasting with ARIMA + ETS + Theta + tree models. Supports covariates — forecast revenue considering ad spend, promotions, and seasonality together

Data quality gate:

• Blocks training when data is insufficient (e.g., "Need 60+ data points for 30-day forecast")
• Warns about missing values, class imbalance, limited data, and time-aggregated data used for classification
• Business-friendly accuracy display: "Predictions within ±8% of actual values" instead of raw ML metrics

Training triggers:

• After KPI collection (automatic retrain when new data arrives)
• On a schedule (daily/weekly)
• Manual (on-demand)

Post-prediction actions:

• "Explore in Queries" — full query page with CSV export, charts, and NL→SQL
• "Get Recommendations" — LLM analyzes prediction results and suggests actionable next steps

Industry use cases:

Industry	Forecast	Classify	Estimate	Detect
E-commerce	Demand forecasting, revenue projections	Customer churn, fraud detection	Customer lifetime value, lead scoring	Unusual transactions, pricing anomalies
Retail	Inventory planning, seasonal demand	Promotion effectiveness, store clustering	Price optimization, basket value	Inventory shrinkage, POS irregularities
SaaS	MRR/ARR forecasting, usage trends	Churn risk, trial-to-paid conversion	Expansion revenue, support ticket priority	Usage spikes, abnormal API patterns
Finance	Cash flow forecasting, market trends	Credit risk, transaction fraud	Portfolio value, risk scoring	Suspicious transactions, market anomalies
Healthcare	Patient volume, resource planning	Readmission risk, diagnosis classification	Treatment cost, length of stay	Abnormal lab results, billing outliers
Manufacturing	Production demand, supply planning	Quality defect detection, equipment failure	Maintenance cost, yield optimization	Sensor anomalies, process deviations
Logistics	Shipment volume, delivery time	Route optimization, delay risk	Freight cost, warehouse capacity	Delivery exceptions, fleet irregularities

CSV & Excel Upload

Upload data files directly into Actyze for instant analysis:

• CSV files with UTF-8 encoding
• Excel files (.xlsx) with single worksheet
• Automatic schema detection
• Custom column type mapping
• Temporary or permanent tables

→ CSV Upload Guide

Role-Based Access Control (RBAC)

Three-tier permission system for secure data access:

• ADMIN: Full system access, user management, schema configuration
• USER: Create dashboards, upload files, query accessible data
• READONLY: View-only access to data intelligence

→ RBAC Guide

Preferred Tables

Mark up to 25 tables (configurable) as preferred. The AI receives full column metadata and descriptions for your preferred tables on every query, prioritizing them over other recommendations.

→ Preferred Tables Guide

Metadata Descriptions

Add business context to your data with organization-level metadata:

• Describe catalogs, schemas, tables, and columns
• Improve AI understanding of your data
• Better query recommendations
• Team-wide knowledge sharing

→ Metadata Guide

Semantic Relationship Graph

The relationship graph stores verified table-to-table JOIN relationships and provides them to the LLM for accurate multi-table queries:

• Convention inference — auto-detects *_id/*_fk column patterns
• Query history mining — extracts proven JOINs from past queries
• Admin curation — manual creation with confidence scoring
• BFS join path resolution — finds optimal paths across tables
• Usage tracking — relationships that get used rank higher over time

Manage via Data Intelligence → Relationships tab. Enabled by default (RELATIONSHIP_GRAPH_ENABLED=true).

Feature Matrix

Feature	USER	ADMIN	READONLY
Query with natural language
View query results
Create/edit dashboards			No
Create scheduled KPIs			No
Create prediction pipelines			No
Upload CSV/Excel files			No
Mark preferred tables			No
Add metadata descriptions			No
Manage table relationships	No	✅	No
User management	No		No
Configure data access	No		No
Manage schema visibility	No		No

Security Features

Data Access Control

• Granular permissions at catalog, schema, and table levels
• User-level and group-level access rules
• Automatic enforcement through Trino access control

Authentication

• JWT-based authentication
• Session management
• Secure password hashing

Audit Logging

• Query history tracking
• User action logging
• Access attempt monitoring

Performance Features

Query Optimization

• Intelligent query caching
• Automatic query rewriting
• Connection pooling

Scalability

• Horizontal pod autoscaling (Kubernetes)
• Load balancing support
• Efficient resource management

Next Steps

Explore individual feature guides:

1. Smart Relationship Graph - How Actyze learns table connections for accurate JOINs
2. Predictive Intelligence - No-code ML predictions (forecast, classify, estimate, detect)
3. Scheduled KPIs - Automated gold-layer materialization
4. Database Connectors - Connect to your data sources
5. CSV Upload - Import data files
6. RBAC - Configure user permissions
7. Preferred Tables - Personalize AI recommendations
8. Metadata - Add business context

Or learn how to configure Actyze:

• LLM Providers - Set up AI models
• Database Connectors - Configure Trino
• Environment Variables - All settings