Dataherald

Hey Product Hunt community! We’re Amir and Anuj from Dataherald. There have been countless game-changing applications of AI, but one area that has remained uncharted is NL-to-SQL. But fear not - we’re changing that. We’ve built a cutting-edge NL-to-SQL engine, meticulously crafted to be the fastest and most accurate in the market, delivering an unmatched developer experience! The Problem Developers struggle to build NL-to-SQL into products because LLMs do not work out-of-the-box. LLMs do not: 1. Utilize metadata and business definitions since they are not already stored in the relational database 2. Conduct complex SQL queries, such as self joins, multiple sub-queries, or rarer operations like windowing, partitioning, lags, temporal calculations, etc. 3. Create NL-to-SQL training dataset(s) for fine-tuned models 4. Assess the accuracy of AI-generated SQL The Solution Dataherald’s API solves these challenges by allowing developers to: - Add business context from various sources (data dictionaries, DBT, Confluence docs, Google docs, etc.) - Identify relevant data tables - Fine-tune NL-to-SQL LLMs (including GPT-4!) specific to their data - Assess confidence levels in AI-generated SQL All this happens with a few simple API calls. Dataherald’s NL-to-SQL API unlocks data access wherever needed: 1. Enabling Complex Data Queries in SaaS Apps: For example a large payroll software company uses Dataherald to embed NL-to-SQL into their front end so that users can ask questions like “How much did I spend on full stack engineers in our European offices in Q2 2023 and how is that trending?” 2. Democratizing Data Insights: Companies integrate our API into Slack, enabling self-serve data requests, and freeing up valuable time for data analysts. 3. Automate Customer Support with Chatbots: Customer support teams dealing with data-heavy queries leverage Dataherald in chatbots, automating inquiries regarding user data. Seamless Integrations: Dataherald is seamlessly integrated with all major data warehouses, including PostgreSQL, Databricks, Snowflake, BigQuery, and DuckDB. Sign up before January 31 to get $500 in credits. Happy heralding :)

Going to give this a try, team Dataherald. Looks interesting.

As a SQL-noob I upvote this! Thanks and goodluck on launch guys ;)