SQLite Analytics Assistant

A read-only natural-language analytics application for small SQLite databases. Unlike the other LLM projects in this profile, this repository focuses on safe structured-data access rather than chat, summarization or retrieval.

Distinct use case

A student, analyst or instructor can ask questions such as:

• Which branch has the highest average marks?
• How many students are in each section?
• Show the top five scores.

The model generates SQL, but deterministic application logic validates the statement before it reaches the database.

Safety architecture

flowchart LR
    A["Natural-language question"] --> B["LLM SQL translation"]
    B --> C["Read-only policy guard"]
    C --> D["SQLite URI in read-only mode"]
    D --> E["Audited SQL and result table"]

Loading

The guard:

• accepts only SELECT and WITH statements
• rejects multiple statements and mutation keywords
• adds a maximum row limit
• opens SQLite with mode=ro
• displays the final SQL for review

Run

git clone https://github.com/sandeep848/langchain-sqlite-assistant.git
cd langchain-sqlite-assistant
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env
streamlit run app.py

Test

pytest

Key files

• app.py — Streamlit analytics workflow
• query_guard.py — deterministic SQL policy and read-only execution
• sqlite.py — sample database generator
• tests/test_query_guard.py — policy regression tests