Skip to main content

Natural Language to SQL / Generative BI / GenBI

Unified NLQ (Natural Language Query)

WrenAI GenBI Architecture

Tools​

Code​

from langchain import OpenAI, SQLDatabase, SQLDatabaseChain

SQLDatabaseChain

Advanced - HLR-SQL​

HLR-SQL, a technique that employs Human-Like Reasoning (HLR) with large language models (LLMs) to translate natural language questions into complex SQL queries. This approach is designed to handle the multi-table joins and intricate logic often found in real-world enterprise databases, which traditional "text-to-SQL" systems struggle with.

Key Concepts of HLR-SQL​

  • Iterative Query Refinement: Unlike conventional methods that generate a single SQL query in one go, HLR-SQL imitates the way a human data analyst works by incrementally composing the final query through a sequence of intermediate steps.
  • Intermediate Sub-queries: The system breaks a complex problem into smaller, manageable sub-questions, generates SQL sub-queries for them, executes these, and stores the results and reasoning steps in a "memory".
  • Self-Correction: By executing intermediate SQL sub-queries and observing the feedback (results or errors) from the database, the LLM agent can identify and correct mistakes or false assumptions made in earlier steps, thus preventing error propagation.
  • Autonomy and Human-in-the-Loop: The HLR-SQL agent can autonomously decide how many iterations are needed to solve a query. It can also be extended to selectively ask a human for help when it encounters significant ambiguity or persistent errors, integrating user feedback into the reasoning process.

Generative BI in Telecom​

Test Prompts​

What are the rental patterns and lifetime value segments of customers, including their geographic clustering and seasonal preferences? answer the above question using the database.

Customer Rental Pattern Analysis - Claude