superduper-io/superduper

Features

• Autonomous AI Agent Frameworks - Provides a comprehensive framework for building self-directed AI agents that coordinate between database storage and machine learning models.
• Database-Integrated AI - Provides the primary architecture for bridging machine learning inference and database operations through standard query interfaces.
• Retrieval Augmented Generation - Functions as a RAG orchestration platform that grounds language model responses in external database context.
• Agent State Persistence - Persists agent memory and computation results in databases to ensure continuity across multi-step workflows.
• Agentic RAG Platforms - Functions as an orchestration platform that combines database lookups with language model prompts for grounded response generation.
• AI Agent Development Toolkits - Ships an integrated toolkit for creating multi-step AI agents that utilize persisted data states for computation.
• AI Agent State Coordination - Coordinates the interaction and execution state between database storage and AI models for autonomous agents.
• AI Workflow Orchestration - Runs configured sequences of components and queries to process data and generate AI-driven results.
• AI Workflow Orchestrators - Implements frameworks for managing multi-step reasoning and operational sequences by connecting AI models with data backends.
• Database Agents - Creates autonomous agents capable of interacting with and querying structured database records to perform knowledge work.
• In-Database Machine Learning - Enables the execution of machine learning training and prediction tasks directly on data hosted within SQL and NoSQL databases.
• LLM Application Frameworks - Offers a framework for building autonomous agents and data-driven applications by coordinating LLMs with database storage.
• Database-Native ML Integration - Provides a database-integrated pipeline for executing machine learning workflows directly within the database storage layer.
• In-Database Model Executions - Executes machine learning training and prediction tasks directly within the database environment to avoid data movement.
• Model Abstractions - Provides unified interfaces for interacting with diverse AI model APIs and local providers, allowing easy swapping.
• Model Training Workflows - Enables attaching training logic to models to refine performance based on persisted database data.
• Self-Hosted AI Models - Enables running large language models and embedding frameworks on private infrastructure for full data control.
• Text Embedding Generators - Transforms text into vector representations using remote APIs to enable semantic search and similarity analysis.
• Sentence Embeddings - Generates sentence-level vector representations using local models to power semantic search.
• Database Backend Integration - Integrates with various SQL and NoSQL database engines to provide persistent storage for AI agents.
• SQL Query Execution - Executes SQL queries against databases with integrated support for complex data types and vector searches.
• Automated Type Inference - Automatically constructs database tables and maps types by analyzing data during record insertion.
• Database Storage Abstractions - Provides a uniform API to interact with diverse SQL and NoSQL databases, abstracting storage engine differences.
• Vector Indexing - Automates the conversion of database outputs into vector representations for semantic search and RAG.
• AI Agent Orchestrators - Ships a backend orchestrator that coordinates model providers and task execution to construct autonomous agents.
• Agent - Parametrizes AI agents and computations into reusable units for consistent deployment across different datasets.
• Private Infrastructure Hosting - Allows full ownership of AI models and data by running all infrastructure on private hardware.
• Functional Application Templates - Enables the creation of parameterized templates that bundle AI logic and configuration for portable deployment.
• Agent Deployment - Provides systems for provisioning and configuring AI agent instances through reusable templates.
• AI Model Integrations - Wraps standard AI models with integration logic to enable their use within database-backed applications.
• Provider Swapping - Allows switching between different AI model providers at runtime to move between cloud APIs and self-hosted models.
• AI Paradigm Implementations - Implements multiple machine learning paradigms, including RAG and generative AI, within a unified framework.
• Chat Model Integrations - Provides standardized interfaces to connect vLLM hosted models into agentic application workflows.
• Data-Driven Model Triggers - Configures AI models to automatically process or respond to new records added to the database.
• Transformer Pipeline Wrappers - Executes machine learning pipelines with custom input and output types to integrate language model capabilities.
• LLM API Integrations - Provides capabilities for connecting to external large language model providers via APIs for agentic workflows.
• Custom Estimator Integrations - Integrates machine learning estimators and tensors with configurable pre-processing and post-processing logic.
• Language Model Fine-Tuning - Provides specialized workflows for adapting pre-trained language models to specific database datasets.
• PyTorch Model Wrappers - Wraps PyTorch models with custom logic to make them compatible with the agent framework.
• PyTorch Training Frameworks - Provides high-level structures and configuration wrappers to organize and execute the training of PyTorch models.
• Scikit-Learn Estimator Wrappers - Provides wrappers for Scikit-Learn estimators to handle data pre-processing and post-processing within ML pipelines.
• OpenAI Model Integrations - Integrates OpenAI models to perform text embeddings, chat completions, and other multimodal tasks.
• Data Abstraction Layers - Organizes data, metadata, and artifacts within a virtual layer that decouples application logic from the physical database.
• Auto-Schema Mapping - Automatically analyzes data types during insertion to construct database tables and map storage types.
• Database Change Tracking - Monitors database inserts, updates, and deletes to trigger automated AI workflows and actions.
• Logic-Integrated Table Management - Integrates database tables into application logic and provides optional schema enforcement.
• Vector Index Synchronization - Provides infrastructure for generating embeddings and maintaining synchronized vector indexes to enable semantic search.
• Vector Similarity Search - Executes similarity searches across stored documents to retrieve relevant context for AI agents based on embedding vectors.
• ML Logic Components - Provides specialized classes for implementing model prediction, training, and quality measurement logic.
• Dynamic Module Loading - Integrates external Python packages and plugins at runtime to extend system capabilities without application restarts.
• Database Change Triggers - Monitors database changes to automatically execute AI models or workflows when records are updated.
• Agent Capability Bundles - Packages pre-configured AI functionality into compact units for simplified shipping and deployment.
• Application Packaging Templates - Packages parameterized application logic and configuration into reusable templates for consistent deployment across datasets.
• Self-Hosted AI Infrastructure - Supports the deployment of large language models and embedding frameworks on private hardware for full control over inference.
• Declarative Configuration Management - Implements a system for defining agent behaviors and component states via database-backed declarative configurations.
• AI Agent Frameworks - End-to-end framework for custom AI applications.
• Mlops Projects - Listed in the “Mlops Projects” section of the Awesome For Beginners awesome list.
• Python Projects - Listed in the “Python Projects” section of the Awesome For Beginners awesome list.

Star history