LLM LangChain Text To SQL Streamlit

(also provided Traditional Chinese version document README-CH.md.)

Use LangChain to integrate with OpenAI or Ollama and connect to a database, enabling the AI to understand the database structure, enhance prompts with Retrieval-Augmented Generation (RAG), achieve Text-to-SQL, and build an interactive frontend web page with Streamlit.

Overview

• Module: langchain-community v0.3.19, langchain-openai v0.3.7, langchain v0.3.20, langchain-ollama v0.2.3
• UI: streamlit v1.42.2

Env

Able to choose OpenAI or Ollama as the LLM and set the environment variables for the selected LLM. If using OpenAI, an API key must be obtained from OpenAI (a paid subscription may be required). Additionally, the embedding model used for RAG can be different from the selected LLM.

# LLM OPENAI or OLLAMA
LLM_TYPE='OLLAMA'

OPENAI_MODEL='gpt-4o-mini'
OPENAI_API_KEY=''

OLLAMA_URL='http://localhost:11434'
OLLAMA_MODEL='gemma3:27b'

# Embedding
OLLAMA_EMBEDDING_URL='http://localhost:11434'
OLLAMA_EMBEDDING_MODEL='llama3.1:70b-instruct-q2_K'

# Database
DB_URL='postgresql://user:password@host:5432/database'

Run

Prerequisites

Modify the prompt in the code based on your database information. Some templates have already been provided.

Install Module

pip install streamlit==1.42.2 pandas==2.2.3 python-dotenv==1.0.1 \
langchain-community==0.3.19 langchain-openai==0.3.7 langchain==0.3.20 \
langchain-ollama==0.2.3

Run

Script Version

python script.py

Web Version

streamlit run app.py

server running at http://localhost:8501

UI

References

• LangChain Official Website