mySQLhelper

Chat with your SQLite databases using a local LLM. Ask questions in plain English, get SQL queries and results. Everything runs offline on your machine.

tinymysqlvid.mp4

Features

Feature	What it does
100% Offline	Your data never leaves your machine. Uses local LLMs via LM Studio
Natural Language	Ask "show me the top 5 customers" and get results
See the SQL	Every query is shown before execution - no black box
Search All Tables	Find a value across your entire database in one click
Forensic Functions	Built-in timestamp converters, Base64/Hex decode, pattern extractors
Interactive Tables	Sort, filter, paginate results. Dark/light theme
Export Sessions	Save your analysis as an HTML report with full audit trail

Note: Designed for localhost use on a single workstation. For network deployment, set SECRET_KEY env var.

Quick Start

1. Get LM Studio and load a model (Qwen, Mistral, etc). Enable the local server on port 1234.

2. Clone and run:

git clone https://github.com/reisset/mysqlhelper.git
cd mysqlhelper
python3 -m venv venv && source venv/bin/activate  # Windows: .\venv\Scripts\activate
pip install -r requirements.txt
flask run

3. Open http://127.0.0.1:5000 and drag in a .db file.

For Forensic Analysts

• Read-only guaranteed - SQL validation + SQLite mode=ro blocks any writes
• Chain of custody - SHA256 hashes logged, timestamped exports
• Audit logs - All queries logged to logs/ with daily rotation
• Air-gapped safe - Zero telemetry, no internet required

Heads up: WAL files (.db-wal, .db-shm) aren't uploaded with the main DB. Checkpoint first if you need recent transactions.

Stack

Flask + Vanilla JS + SQLite + Local LLM (OpenAI-compatible API)

License

MIT

---

Built by Reisset