mnemiq vs Natural Language SQL

mnemiq is an open-source text-to-SQL engine for enterprise data agents, configured and measured against a user's own database. Before execution, a deterministic check verifies that proposed SQL is read-only, uses permitted objects, compiles in the source dialect, and passes EXPLAIN. Access permissions apply before schema retrieval, so the model is not shown tables the caller cannot access. Each answer includes the SQL run, the tables it touched, and the enrichment version used. mnemiq can profile columns and build descriptions, grain, glossary terms, and coded-value meanings; a human owner can certify semantic definitions. Listed databases are PostgreSQL, SQLite, DuckDB, Oracle, Snowflake, and Databricks, with DuckDB as the universal executor. Models can use any OpenAI-compatible /v1 endpoint, including local servers such as vLLM, Ollama, llama.cpp server, and LM Studio. Its MCP server offers two read-only, access-scoped tools: db_read(question) and get_schema(). One process can also serve a workbench and HTTP endpoints for asking, chatting, and retrieving schema. The Apache-2.0 engine is free and runs in the user's environment. Local model hosting keeps schemas, questions, and rows inside the user's network, according to the maker.

Natural Language SQL turns plain-English questions about databases into SQL, validates and runs the query, then streams back results. It connects to PostgreSQL and MySQL, including multiple databases of the same type in parallel. Hybrid retrieval combines BM25 and sentence-transformer vectors to select schema context for a question. Users can review and edit generated SQL before execution, while the interface shows progress through query stages. Results can be sorted, paginated at 50 rows per page, downloaded as CSV or JSON, or copied as tab-separated text. SELECT queries receive a default limit, explicit result counts are capped at 1,000 rows, and execution times out after 30 seconds by default. INSERT, UPDATE and DELETE are allowed by default, with warning banners; the project recommends a read-only database user to block writes. Ollama runs locally by default, keeping questions, schema and generated SQL on the user’s machine. OpenAI, Gemini and Groq are optional cloud providers and receive the question, relevant schema portion and SQL context. The project is free under GPL-3.0. Its recommended Docker setup needs Docker 20.10 or later with Compose v2 and about 10 GB of free disk space.