Let Your LLM
Fetch Data from Databases

Please reach out to us at db-ally

In contrast to available solutions, db-ally generates Intermediate Query Language (IQL) rather than directly producing SQL queries. This approach offers several significant benefits: Faster & Cheaper LLM Responses: Using IQL requires fewer tokens overall, resulting in more efficient and cost-effective responses from large language models (LLMs). Technology Agnostic: IQL can be transformed into a wide selection of data sources, providing flexibility and compatibility with various database systems. Reliability & Security: End-users have full control over the executed queries, ensuring greater security and reducing the risk of executing malicious SQL payloads. Easier to Work with for LLMs: Complex domain knowledge can be embedded in IQL methods, making it simpler for LLMs to comprehend and generate accurate responses. By leveraging IQL, db-ally delivers a robust, versatile, and secure solution for querying structured data using natural language.

At its core, db-ally is designed to be agnostic to any specific technology, offering a wide selection of integrations to meet diverse needs. Our flexible framework ensures that integrating additional technologies is straightforward for any third party. Here is the initial scope of supported technologies: Data Source: PostgreSQL: Our initial integration supports PostgreSQL as the primary data source, ensuring robust and reliable data management. Large Language Models (LLMs): OpenAI & Anyscale Endpoints: db-ally supports these LLMs, enabling powerful natural language processing capabilities. Vector Search Capabilities: Meta FAISS: For efficient and scalable vector search, db-ally integrates with Meta FAISS. Observability: Langchain’s Langsmith: To ensure comprehensive observability and monitoring, we utilize Langsmith from Langchain. This modular and technology-agnostic design allows db-ally to easily incorporate additional technologies, making it a versatile and future-proof solution for your data querying needs.

db-ally excels in embedding complex domain-specific knowledge into its Intermediate Query Language (IQL), which offers several benefits over traditional Text2SQL solutions: Simplified Representation: IQL encapsulates complex logic into easier methods for LLMs to understand and work with, reducing the complexity of natural language queries. Custom Methods: Users can define custom methods within IQL to handle specific domain requirements, ensuring the generated queries are accurate and contextually relevant. Enhanced LLM Understanding: IQL allows LLMs to generate more precise and relevant responses by providing a structured and simplified representation of domain knowledge, improving overall query accuracy. These features make db-ally particularly effective in scenarios requiring detailed domain-specific knowledge, where traditional Text2SQL solutions might struggle.

Yes, db-ally is designed flexibly, allowing easy integration with various data sources and technologies. While our initial scope includes PostgreSQL, OpenAI, Anyscale Endpoints, Meta FAISS, and Langchain’s Langsmith, our technology-agnostic approach ensures that third parties can integrate additional data sources and technologies as needed. This makes db-ally a versatile solution capable of adapting to various applications and environments.

While Text2SQL solutions directly generate SQL queries from natural language inputs, db-ally uses an Intermediate Query Language (IQL) to provide several advantages: Security: IQL acts as a protective layer, reducing the risk of SQL injection attacks and other security vulnerabilities common in direct Text2SQL approaches. Reliability: db-ally’s IQL ensures more consistent and reliable query generation, avoiding the unpredictability often seen with Text2SQL methods. Performance: By reducing the token count needed for LLM responses, db-ally offers faster and more cost-effective performance than traditional Text2SQL systems. Complex Query Handling: IQL simplifies the representation of complex domain-specific logic, making it easier for LLMs to generate accurate responses without getting bogged down by intricate SQL syntax.