orionbelt-semantic-layer

Overview

OrionBelt Semantic Layer is an API-first semantic engine that compiles declarative YAML models into optimized, dialect-specific SQL across multiple database platforms including BigQuery, PostgreSQL, Snowflake, ClickHouse, Dremio, Databricks, DuckDB, and MySQL. It enables AI agents to query databases using business concepts rather than raw SQL through a REST API and MCP integration. The solution provides analytics as code with version-controlled YAML models, dialect-specific SQL generation, and a complete integration surface including UI, API, and MCP server.

Try asking AI

After installing, here are 6 things you can ask your AI assistant:

• compile_query

  Compile a query into SQL based on the semantic model

  Try:Calculate total revenue by countryTry:Show me the top 5 products by salesTry:What's the monthly trend of user registrations?
• validate_model

  Validate a semantic model definition

  Try:Check if my model definition is validTry:Validate the sales model for errorsTry:Is this OBML model correctly structured?
• list_models

  List available semantic models

  Try:Show me all available modelsTry:What models are loaded in the system?Try:List the marketing analytics models
• query_plan

  Get query plan information without executing the query

  Try:Show me the query plan for this requestTry:How would you join these tables?Try:Explain the join strategy for multi-fact query
• find_semantics

  Find semantic elements in the model

  Try:Find all dimensions related to timeTry:Search for revenue metricsTry:What measures are available for product data?
• execute_query

  Execute a compiled query against the database

  Try:Run this SQL query against BigQueryTry:Execute the customer analysis queryTry:Get results for the monthly sales report
• compile_examples

  Get example queries from model definitions

  Try:Show me examples from the sales modelTry:What queries are available in the marketing model?Try:Get example use cases for the financial model
• heartbeat

  Invalidate cache for dependent queries

  Try:Mark customer table as updatedTry:Refresh cache for sales dataTry:Invalidate all queries depending on orders
• get_model_health

  Check model health including orphaned elements

  Try:Check the health of the sales modelTry:Are there any fan-traps in my model?Try:Report on model integrity issues
• export_model

  Export model in different formats

  Try:Export the model as RDF/OWLTry:Convert OBML to OSI formatTry:Export the model schema as graph
• get_warnings

  Get warnings from model operations

  Try:Show me warnings from the last compilationTry:Are there any issues with my model?Try:Report on validation warnings
• discover_databases

  Discover available databases and schemas

  Try:What databases are available?Try:List all schemas in the data warehouseTry:Show me the customer data sources

Comparable tools

Installation

pip install orionbelt-semantic-layer-mcp

{
  "mcpServers": {
    "orionbelt": {
      "command": "uvx",
      "args": ["orionbelt-semantic-layer-mcp"]
    }
  }
}

pip install orionbelt-semantic-layer

docker run -p 8080:8080 ralforion/orionbelt-api

FAQ

What databases does OrionBelt support?

OrionBelt supports BigQuery, PostgreSQL, Snowflake, ClickHouse, Dremio, Databricks, DuckDB/MotherDuck, and MySQL.

How does OrionBelt handle SQL injection prevention?

OrionBelt uses AST-based SQL generation, not string templates, which ensures injection-safe SQL generation across all supported dialects.

Can I use OrionBelt with other AI assistants besides Claude?

Yes, the MCP server also works with Copilot, Cursor, and Windsurf. The main API can be integrated with any platform that supports REST or HTTP endpoints.

Compare orionbelt-semantic-layer with