PromptQL - A data access agent with out-of-the-box connectors and agentic query planning.

PromptQL A data access agent with out-of-the-box connectors and agentic query planning.

What is PromptQL?

PromptQL facilitates advanced data interaction for AI systems by providing seamless integration with various data sources through its connectors. It supports SQL databases, NoSQL stores, APIs, and SaaS platforms, unifying them under a single schema for AI access. The system employs dynamic query planning, allowing plans to be adjusted during execution based on intermediate results or user instructions, moving beyond rigid retrieval methods.

This tool features a programmatic runtime environment enabling complex computations, data transformations, and statistical analyses, even on large datasets processed outside the LLM's context window. It supports actions and mutations, allowing AI to update data, call APIs, or invoke functions. PromptQL incorporates structured memory to overcome LLM context limitations and provides explainability by detailing execution steps. Integrated fine-grained authorization ensures secure, role-based data access control.

Features

Dynamic Query Planning: Builds adaptable query plans modified during execution based on results or instructions.
Connectors: Integrates seamlessly with diverse data sources (SQL, NoSQL, APIs, SaaS) under a unified schema.
Computation: Executes complex calculations, transformations, and analyses, handling large datasets outside the LLM context.
Actions & Mutations: Enables AI to update data, call APIs, or invoke functions for task completion.
Structured Memory: Stores information outside the LLM context window to manage large data and maintain conversation context.
Explainability: Provides detailed execution plans and explanations for query steps.
Fine-Grained Authorization: Implements role-based access control (row, column, object level) for secure data querying.

Use Cases

Analyzing sales performance across multiple data sources.
Generating customer reports combining purchase history, support tickets, and interactions.
Calculating complex business metrics like Customer Lifetime Value (CLTV).
Automating actions based on data analysis (e.g., sending reminders, updating records).
Projecting future performance based on historical data analysis.
Building AI assistants for specific domains like engineering management or customer support.
Creating natural language interfaces for exploring complex datasets.

FAQs

How does PromptQL handle diverse data sources?

It uses connectors to seamlessly integrate with various sources like SQL databases, NoSQL stores, APIs, and SaaS platforms under a unified schema.
Can PromptQL perform actions beyond data retrieval?

Yes, it enables AI to perform actions and mutations, such as updating data records, calling external APIs, or invoking specific functions for end-to-end task completion.
How does PromptQL address the context window limitations of LLMs?

It utilizes structured memory artifacts to store and retrieve information outside the LLM's context window, ensuring consistency and enabling the handling of large datasets.
Is data access secure when using PromptQL?

Yes, PromptQL integrates fine-grained, role-based access control directly into the querying process, enforcing predefined authorization policies at row, column, and object levels.
How does PromptQL provide transparency in its operations?

It offers explainability by providing detailed execution plans and allowing users to explicitly ask for explanations of each step taken during data processing or query execution.