Actyze

From question to insight in seconds

Ask questions in plain English, get SQL and visualizations instantly. Train ML models without writing code. Monitor KPIs on autopilot.

Core Platform

AI-Powered Query Engine

Ask a question in natural language—Actyze writes the SQL, runs it across your databases, and visualizes the result. No SQL knowledge required.

Voice & Text Input

Type or speak your question. Actyze supports 50+ languages and converts your intent into optimized SQL.

Smart Intent Detection

ML-based intent classification understands whether you want a new query, a refinement, or a correction—no redundant LLM calls.

Federated Queries

Query across PostgreSQL, MySQL, SQL Server, Oracle, Snowflake, and MongoDB in a single question via Trino federation.

New

Predictive Intelligence

Build ML prediction pipelines from your data without writing a line of code. Choose an outcome, pick your data, and Actyze handles model selection, training, and deployment.

Forecast

Predict future values—revenue, demand, traffic. Powered by AutoGluon and XGBoost time-series models.

Classify

Predict categories—churn risk, fraud detection, lead scoring. LightGBM or XGBoost selected automatically.

Estimate

Predict continuous values—customer lifetime value, pricing, scoring. Regression models handle the math.

Detect Anomalies

Surface unusual data points with Isolation Forest. No labels needed—fully unsupervised anomaly detection.

How it works

1

Choose an outcome

Pick forecast, classify, estimate, or detect. Select your data source—a KPI or custom SQL.

2

Actyze trains the model

Data quality checks run first. Then the best model (XGBoost, LightGBM, or AutoGluon) trains automatically.

3

Query predictions naturally

Results land in a queryable table. Ask “show churn predictions” and the AI finds them automatically.

New

Scheduled KPIs

Define a SQL query, set a schedule, and Actyze materializes the results into a gold-layer table automatically. No ETL pipelines to build or maintain.

Automatic Collection

Schedule KPI collection every 1–24 hours. Data appends with timestamps, building a time-series automatically.

AI-Discoverable

Materialized KPI tables register with the semantic layer. Ask “show daily revenue trends” and the AI finds the right table.

Prediction Triggers

Link a KPI to a prediction pipeline. When fresh data arrives, ML models retrain automatically on the latest numbers.

Core Differentiator

Smart Semantic Layer

Most AI SQL tools guess how tables connect. Actyze builds a living graph of your table relationships—learning from naming conventions, query history, and admin input—so the AI writes accurate JOINs instead of hallucinating them.

Inferred Relationships

Scans column names for foreign-key patterns (e.g. customer_idcustomers.id) and creates relationships automatically on first schema load.

Mined from Query History

Parses every successful query with sqlglot to extract real JOIN patterns. Confidence grows with frequency—the more a join succeeds, the higher it scores.

Admin-Verified

Admins can create, verify, or disable relationships through the UI. Verified relationships get the highest confidence and override inferred or mined ones.

How the graph powers every query

1

Direct relationships

If two tables are directly connected (e.g. orders → customers), the AI gets the exact join condition with confidence score.

2

Multi-hop path finding

For indirect connections, BFS traversal finds the optimal path (e.g. orders → customers → nations) and passes the full chain to the LLM.

3

Schema fallback

If no relationship exists yet, the LLM falls back to full schema context. As queries succeed, the graph learns the new pattern automatically.

Confidence Scoring

Every relationship carries a 0–1 confidence score. Mined relationships gain confidence with each successful use. Low-confidence joins are deprioritized.

Relationship Types

Tracks 1:1, 1:N, N:1, and M:N relationship types so the LLM knows when to expect multiple rows or unique matches.

Usage Tracking

See which relationships power your AI-generated queries. Usage count and last-used timestamps help you audit what the AI relies on.

Bulk Import

Bootstrap the graph from CSV or add relationships one-by-one through the admin editor. Disable problematic joins without deleting them.

Infrastructure

Self-Hosted & Secure

Actyze runs entirely in your infrastructure. Your data, queries, and LLM calls never leave your network.

BYOC

Deploy with Docker Compose or Helm. Your cloud, your rules.

Bring Your Own LLM

Works with Claude, GPT-4, or any provider via LiteLLM. Use your own API key.

AGPL-3.0 Open Source

Full source code available. Every feature, free, no limits.

Enterprise Ready

SSO, RBAC, audit logging, air-gapped deployments for compliance-heavy organizations.

Ready to get started?

Deploy Actyze for free with Docker Compose. All features included, no limits.