Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FinAI Data Assistant: LLM-based Financial Database Query Processing with the OpenAI Function Calling API

Juhyeong Kim, Yejin Kim, Youngbin Lee, Hyunwoo Byun

TL;DR

FinAI Data Assistant presents a pragmatic NL-to-database approach that pairs an LLM with the OpenAI Function Calling API and a compact library of parameterized SQL templates to enable reliable, low-latency access to financial data. The method emphasizes deterministic execution, auditability, and cost control over full SQL generation, achieving superior end-to-end performance compared with text-to-SQL baselines. Empirical results show that LLMs alone struggle with time-sensitive data and exhibit look-ahead bias for prices, ticker mapping is highly reliable for major indices, and the function-calling approach delivers faster, cheaper, and more dependable query processing. The work underscores practical deployment considerations and highlights opportunities for safe, scalable production use in financial analytics contexts.

Abstract

We present FinAI Data Assistant, a practical approach for natural-language querying over financial databases that combines large language models (LLMs) with the OpenAI Function Calling API. Rather than synthesizing complete SQL via text-to-SQL, our system routes user requests to a small library of vetted, parameterized queries, trading generative flexibility for reliability, low latency, and cost efficiency. We empirically study three questions: (RQ1) whether LLMs alone can reliably recall or extrapolate time-dependent financial data without external retrieval; (RQ2) how well LLMs map company names to stock ticker symbols; and (RQ3) whether function calling outperforms text-to-SQL for end-to-end database query processing. Across controlled experiments on prices and fundamentals, LLM-only predictions exhibit non-negligible error and show look-ahead bias primarily for stock prices relative to model knowledge cutoffs. Ticker-mapping accuracy is near-perfect for NASDAQ-100 constituents and high for S\&P~500 firms. Finally, FinAI Data Assistant achieves lower latency and cost and higher reliability than a text-to-SQL baseline on our task suite. We discuss design trade-offs, limitations, and avenues for deployment.

FinAI Data Assistant: LLM-based Financial Database Query Processing with the OpenAI Function Calling API

TL;DR

FinAI Data Assistant presents a pragmatic NL-to-database approach that pairs an LLM with the OpenAI Function Calling API and a compact library of parameterized SQL templates to enable reliable, low-latency access to financial data. The method emphasizes deterministic execution, auditability, and cost control over full SQL generation, achieving superior end-to-end performance compared with text-to-SQL baselines. Empirical results show that LLMs alone struggle with time-sensitive data and exhibit look-ahead bias for prices, ticker mapping is highly reliable for major indices, and the function-calling approach delivers faster, cheaper, and more dependable query processing. The work underscores practical deployment considerations and highlights opportunities for safe, scalable production use in financial analytics contexts.

Abstract

We present FinAI Data Assistant, a practical approach for natural-language querying over financial databases that combines large language models (LLMs) with the OpenAI Function Calling API. Rather than synthesizing complete SQL via text-to-SQL, our system routes user requests to a small library of vetted, parameterized queries, trading generative flexibility for reliability, low latency, and cost efficiency. We empirically study three questions: (RQ1) whether LLMs alone can reliably recall or extrapolate time-dependent financial data without external retrieval; (RQ2) how well LLMs map company names to stock ticker symbols; and (RQ3) whether function calling outperforms text-to-SQL for end-to-end database query processing. Across controlled experiments on prices and fundamentals, LLM-only predictions exhibit non-negligible error and show look-ahead bias primarily for stock prices relative to model knowledge cutoffs. Ticker-mapping accuracy is near-perfect for NASDAQ-100 constituents and high for S\&P~500 firms. Finally, FinAI Data Assistant achieves lower latency and cost and higher reliability than a text-to-SQL baseline on our task suite. We discuss design trade-offs, limitations, and avenues for deployment.
Paper Structure (11 sections, 5 equations, 2 figures, 2 tables)

This paper contains 11 sections, 5 equations, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Methods
  4. Preliminaries: OpenAI Function Calling API
  5. FinAI Data Assistant
  6. Experiments
  7. Necessity of LLM--Database Integration (RQ1)
  8. Basic Company Information Understanding (RQ2)
  9. Database Query Performance (RQ3)
  10. Conclusion
  11. Test cases for Database Query Performance (RQ3) experiment.

Figures (2)

  • Figure 1: Ticker coverage test. We provide a company name and ask the model for the corresponding ticker. "Latest" denotes accuracy for the most recent index constituents. "Knowledge Cutoff" denotes accuracy for the same sets restricted to membership before each model's knowledge cutoff.
  • Figure 2: Cost, time, and accuracy for database query processing: Function Calling (FinAI Data Assistant) versus text-to-SQL. All metrics are measured for successful completion of each test case. Error bars in the cost and time plots indicate $\pm 1$ standard deviation.