ChatDB Magic: Augmenting LLMs with Databases as Their Symbolic Memory

Refining ChatDB with SQLite database support and improved convenience

Forked from https://github.com/huchenxucs/ChatDB

Modifications:

• Added support for SQLite and MySQL.
• Default database transitioned to SQLite; you can switch back to MySQL by modifying the .env file.

  DB_TYPE=sqlite or DB_TYPE=mysql

• Implemented Answer Analysis, utilizing ChatGPT to analyze SQL results.

  Optional, set use_semantic_answer=False in chatdb.py to disable it.

  USER INPUT: Doctor's nameDatabase response:+---------------------+| name |+---------------------+| Alice Shaw, MD, PhD |+---------------------+Answer: The name of the doctor is Alice Shaw, MD, PhD.

• Introduced automatic language translation for Answer Analysis.
• Revised prompt template to JSON format and integrated colored text into console chat.
• Added support for the gpt-3.5-turbo-16k model, set FAST_LLM_MODEL=gpt-3.5-turbo-16k in .env file.
• Included configuration for adding synonyms for table fields in the Prompt, set synonyms in chatdb_prompts.py file.

Run and Use:

• Follow the instructions in the README.

• Quick demo:

   copy .env.template .env # Set your OpenAI key and host, if use mysql, need to set host, port, name, password.
  pip install -r requirements.txt
  python chatdb.py
  START!
  USER INPUT: How many apples does the customer named 'Chenzhuang Du' bought on 2010-03-27
  Step1: Retrieve the quantity of apples bought by the customer
  SELECT si.quantity_sold
  FROM customers c
  JOIN sales s ON c.customer_id = s.customer_id
  JOIN sale_items si ON s.sale_id = si.sale_id
  JOIN fruits f ON si.fruit_id = f.fruit_id
  WHERE c.first_name = 'Chenzhuang' AND c.last_name = 'Du' AND f.fruit_name = 'apple' AND s.sale_date = '2010-03-27';
  Execute: Database response:
  no results found.
  Finish!

• Use your own data:

   I fixed some bugs, it's better to follow these steps:
  1. Put your sqlite db file to replace try1024.db. 2. Open chatdb.py, set init_db = True to False, to disable create and insert sample data.
  3. Open table_schema.py and add your table schema and then add these variables to `tables` list at the bottom.
  4. Run chatdb.py
  5. Input your query, like the demo above.
  Also, you can change sql_examples.py to use your own examples.
  

• Demo