Open-source Agentic BI platform that defines governed metrics once in a context layer and serves dashboards, AI agents, and data apps from it.
- Stars
- 5,139
- Forks
- 1,022
- Open issues
- 334
- Last commit
- 7 days ago
What is it?
SuperSonic is an open-source platform for analytics that pairs a natural-language query interface, where business users ask questions and receive charts, with a headless semantic layer, where analytics engineers define metrics, dimensions and their relationships. The semantic model serves both paths: the LLM-based Text2SQL pipeline uses it for context retrieval to reduce hallucinated fields, and the same model is exposed through a Headless BI query API.
The pipeline is split into components: a knowledge base that indexes schema elements, a schema mapper that matches query terms against it, semantic parsers and correctors that combine rule-based and LLM-based logic, and a translator that produces executable SQL. Chat plugins for third-party tools and chat memory for few-shot recall are also included. Access control covers dataset, column and row levels, and custom components can be added through Java SPI.
The project is written in Java and can be run with Docker Compose or from a prebuilt binary that starts as a standalone service.
Key features
- 01Natural-language querying with automatic chart visualization for business users.
- 02Headless semantic layer for defining metrics, dimensions and entity relationships.
- 03Hybrid rule-based and LLM-based semantic parsing and correction.
- 04Dataset, column and row-level access control.
- 05Multi-turn conversation, input auto-completion and post-query recommendations.
- 06Extensible through Java SPI and third-party chat plugins.
Use cases
- Letting non-technical staff ask questions of company data in plain language.
- Building a governed semantic model that both dashboards and an LLM query against.
- Adding a natural-language API on top of an existing metrics layer.
- Prototyping Text2SQL pipelines with rule-based and LLM components.
Topics
FAQ
- What is supersonic?
- An AI+BI platform that combines LLM-based natural language querying (Chat BI) with a semantic-layer-driven Headless BI interface. SuperSonic is an open-source platform for analytics that pairs a natural-language query interface, where business users ask questions and receive charts, with a headless semantic layer, where analytics engineers define metrics, dimensions and their relationships. The semantic model serves both paths: the LLM-based Text2SQL pipeline uses it for context retrieval to reduce hallucinated fields, and the same model is exposed through a Headless BI query API.
- Is supersonic free to use? What license does it have?
- supersonic is open source under the Other license. Check the license terms before using it in commercial products.
- Is supersonic still maintained?
- Actively maintained. The last commit to supersonic was on Oct 15, 2026.
- What can I use supersonic for?
- Letting non-technical staff ask questions of company data in plain language. Building a governed semantic model that both dashboards and an LLM query against. Adding a natural-language API on top of an existing metrics layer. Prototyping Text2SQL pipelines with rule-based and LLM components.
- Where can I get supersonic?
- The source code and installation instructions are on GitHub: https://github.com/tencentmusic/supersonic
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
Open-source ChatBI system that turns natural-language questions into SQL queries over your databases, using LLMs and RAG.
Open-source GenBI engine that lets AI agents turn natural-language questions into governed SQL and shareable dashboards over 20+ data sources.
Open-source business intelligence tool from Frappe for building queries, charts and dashboards without writing SQL.
Open-source, self-hosted tool for building SQL-based metrics and shareable documents and dashboards for teams.
BI as code: define data sources, a semantic layer and dashboards in YAML, with local dev server, PR review and an MCP server for AI agents.