DecisionBox
DecisionBox is an open-source, self-hostable platform that runs an autonomous AI agent on your data warehouse to discover insights and recommendations. It connects to major warehouses, writes and executes SQL, validates findings with independent queries, and logs every step for transparency. It is for teams that make data-driven decisions and want automated, trustworthy analysis without writing queries.
✨ Key features
- Autonomous agent loop writes and executes SQL
- Independent verification queries validate every insight
- Full transparency: SQL, reasoning, and decisions logged
- Supports BigQuery, Redshift, Snowflake, PostgreSQL, Databricks, MSSQL
- Works with Claude, OpenAI, Ollama, Vertex AI, Bedrock, Azure AI
- Domain packs for e-commerce, social, gaming; dynamic packs
🎯 Use cases
- E-commerce: detect cart abandonment spikes and recommend free shipping thresholds
- SaaS: identify features that reduce churn and suggest onboarding improvements
- Fraud detection: flag patterns leading to chargebacks and prevent losses
- SQL performance: find slow queries and reduce warehouse compute costs
- Gaming: analyze level failure rates to improve retention
📦 Installation
🧰 Requirements: Requires Docker and Docker Compose for quick start; a data warehouse connection and an LLM API key (Anthropic, OpenAI, or cloud providers) are needed.
git clone https://github.com/decisionbox-io/decisionbox-platform.git
cd decisionbox-platform
docker compose up -d
open http://localhost:3000
🚀 Usage
After installation, open the dashboard at http://localhost:3000 and follow the setup wizard to create a project, connect your warehouse, and configure an LLM provider. Then run a discovery to let the agent explore your data and review insights.
❓ FAQ
What data warehouses does DecisionBox support?
It supports BigQuery, Redshift, Snowflake, PostgreSQL, Databricks, and Microsoft SQL Server, with more shipping weekly.
How does DecisionBox ensure insights are accurate?
Every insight is re-queried against your warehouse with independent verification queries before it is delivered, preventing hallucinations.
Can I use local LLMs?
Yes, DecisionBox supports Ollama for local models, as well as cloud providers like Claude, OpenAI, Vertex AI, Bedrock, and Azure AI.
Is DecisionBox open source?
Yes, it is open source under AGPL v3 and self-hostable, with Docker Compose, Helm charts, and Terraform modules for deployment.
📊 Repository
🤖 Overview, features, install steps and FAQ were generated from the project's README on Sep 4, 2026. Always check the original source before running commands.