Natural Language SQL

Summary

Natural Language SQL turns plain-English database questions into SQL, validates and runs the query, then streams results back. It connects to PostgreSQL and MySQL, including multiple databases of the same type at once. Hybrid retrieval uses BM25 and sentence-transformer vectors to choose relevant schema context. Before execution, users can review and edit the generated SQL, 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 machine, and the project says it has no telemetry. OpenAI, Gemini, and Groq are optional cloud providers; those providers 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+ with Compose v2, about 10 GB of disk space, and downloads an approximately 5 GB model on first run.

Who it is for

It suits users who want to ask questions of PostgreSQL or MySQL databases in plain English and inspect SQL before running it. Local Ollama is a fit for those who want questions, schema, and generated SQL to remain on their machine.

Compared on AI SQL generators

Query explanations

Yesgithub.com

Deployment

self_hostedgithub.com

Schema context

Yesgithub.com

Facts

Purpose

The engine turns plain-English database questions into SQL, validates and runs the SQL, then streams back results.github.com · 4 Oct 2026

AI providers

Ollama runs locally by default, with OpenAI, Gemini, and Groq available as opt-in cloud providers.github.com · 4 Oct 2026

Database support

It connects to PostgreSQL and MySQL and can query multiple same-type databases in parallel.github.com · 4 Oct 2026

Schema retrieval

Hybrid retrieval combines BM25 and sentence-transformer vectors to select relevant schema context for a question.github.com · 4 Oct 2026

Query workflow

Users can review and edit generated SQL before execution, and the interface streams progress through the query stages.github.com · 4 Oct 2026

Results

Results support sorting, pagination at 50 rows per page, CSV and JSON downloads, and tab-separated copying.github.com · 4 Oct 2026

Query limits

SELECT queries receive a default LIMIT, explicit result counts are capped at 1,000 rows by default, and query execution times out after 30 seconds by default.github.com · 4 Oct 2026

Write queries

INSERT, UPDATE, and DELETE are allowed by default, with warning banners; the README recommends a read-only database user to block writes.github.com · 4 Oct 2026

Local privacy

With the default Ollama provider, questions, schema, and generated SQL stay on the user's machine, and the app says it has no telemetry.github.com · 4 Oct 2026

Cloud privacy

With an opt-in cloud provider, the question, relevant schema portion, and generated SQL context are sent to that provider; database passwords and query result rows are excluded.github.com · 4 Oct 2026

Credential security

Database passwords and cloud API keys are Fernet-encrypted at rest, redacted from logs, and masked in API responses.github.com · 4 Oct 2026

Deployment requirements

The recommended Docker setup requires Docker 20.10+ with Compose v2 and about 10 GB of free disk space; the first run downloads an approximately 5 GB model.github.com · 4 Oct 2026

License

The repository lists the project under the GPL-3.0 license.github.com · 4 Oct 2026

Local AI

Ollama runs locally by default, and the project says questions, database schema, and generated SQL stay on the user's machine with this provider.github.com · 5 Oct 2026

Cloud model options

OpenAI, Google Gemini, and Groq are opt-in providers; using one sends the question, relevant schema portion, and SQL context to that provider.github.com · 5 Oct 2026

Database integrations

The app supports connecting to PostgreSQL and MySQL databases, including multiple databases at once.github.com · 5 Oct 2026

Privacy

The project says it collects no telemetry; query results, schema metadata, and query history are held in memory for the session and are not written to disk by the application.github.com · 5 Oct 2026

Benchmark

The README reports 52.32% execution accuracy on Spider dev.github.com · 5 Oct 2026

Where it ranks on Everything Xiaomi

• Best AI SQL Generators in 2026#1 of 19

Sources

• github.com/jmanoj0905/natural-language-sql· checked 4 Oct 2026
• github.com/jmanoj0905/natural-language-sql/blob/ma· checked 4 Oct 2026