DataHerald

Dataherald is a software company that provides a natural language-to-SQL API, enabling developers to integrate text-to-SQL capabilities into their products. The Dataherald engine can be deployed within minutes using a hosted API, allowing users to interact with SQL databases using natural language queries.

Key features:
- Natural Language-to-SQL API: Converts natural language queries into SQL commands
- Custom Agents and Fine-Tuning: Combines custom agents with fine-tuning and built-in evaluation for enhanced performance
- Semantic Layer Creation: Injects business context directly into tables, columns, or entire databases
- Built-in Evaluator: Monitors the model's performance over time and enables feedback learning
- Usage-Based Pricing: Charges self-serve users based on their actual usage without minimums or plan fees
- BYOC (Bring Your Own Code): Ensures seamless integration into existing data stacks
- Synthetic Data Generation: Enhances agent performance through fine-tuning on synthetic data

How it works:
1. Configure the Dataherald Engine: Set up the engine and deploy it in your application
2. Add Business Context: Add instructions directly to your database resources to capture unique business contexts
3. Monitor and Fine-Tune: Observe every query, model, and fine-tuning process through the admin console for continuous improvement

Integrations:
Dataherald supports integration with various data stacks, including BigQuery, PostgreSQL, Databricks, and Snowflake.

Use of AI:
Dataherald leverages generative artificial intelligence through its support for fine-tuning on GPT 3.5 and GPT 4 models, delivering accurate and fast text-to-SQL translations tailored to specific business needs and contexts.

AI foundation model:
Dataherald supports fine-tuning on GPT 3.5 and GPT 4 models.

How to access:
Dataherald is primarily available as a web app, offering a hosted API that developers can integrate into their applications with just a few lines of code.