Gen AI Builder ============== .. _gen-ai-builder-1: Gen AI Builder ============== Gen AI Builder provides comprehensive tooling for developing intelligent applications that leverage organizational knowledge through :ref:`retrieval‑augmented generation ` patterns. The platform operates within controlled infrastructure environments, ensuring complete organizational sovereignty over data, models, and inference operations. New here? Start with :ref:`Gen AI Builder Hub ` , review :ref:`Gen AI Builder Concepts (Assistants, Knowledge, Tools, Threads) ` and the :ref:`Gen AI Architecture (Entities and Flows) ` , then use the :ref:`Gen AI How-To Guides ` or jump into the :ref:`Customer Service Assistant Implementation Guide ` . Platform Architecture --------------------- Gen AI Builder integrates multiple components to deliver end-to-end AI application development capabilities within the EDB Postgres AI ecosystem. The platform emphasizes data sovereignty by maintaining all operations within organizational boundaries while providing enterprise-grade governance and auditability. Explore the :ref:`Assistants Reference Manual ` , :ref:`Knowledge Bases Reference Manual ` , :ref:`Retrievers Reference Manual ` , :ref:`Rulesets Manual ` , :ref:`Structures Reference Manual ` , :ref:`Tools Reference Manual ` , :ref:`Threads Reference Manual ` , :ref:`Configure Data Sources in Gen AI Builder ` , :ref:`Data Lake Reference Manual ` , and :ref:`LibrariesΩ Manual ` manuals. Core Infrastructure Integration ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ **Hybrid Manager Foundation** Gen AI Builder operates within :ref:`EDB’s vision for AI with Postgres and Hybrid Manager ` ’s Kubernetes orchestration layer, leveraging distributed computing capabilities while maintaining centralized governance. This architecture ensures consistent operational procedures across development and production environments. **Data Estate Integration** :ref:`Knowledge Bases Reference Manual ` connect directly to Postgres databases and object storage systems (see :ref:`Data Lake Reference Manual ` ) within organizational infrastructure, eliminating external data dependencies while supporting diverse :ref:`Configure Data Sources in Gen AI Builder ` types. **Model Serving Integration** Applications access language models and embedding models through the :ref:`AI Factory Models ` infrastructure, ensuring consistent governance policies across all AI capabilities. Architectural Components ------------------------ Knowledge Base Framework ^^^^^^^^^^^^^^^^^^^^^^^^ :ref:`Knowledge Bases Reference Manual ` serve as the foundation for retrieval‑augmented generation applications by combining structured database content with unstructured document repositories. The framework provides semantic search capabilities through vector embeddings (see :ref:`Vector Engine (PGvector) ` ) while maintaining complete data lineage. **Data Source Integration** - Postgres database tables with semantic indexing - Document repositories from object storage systems - Real-time data feeds through API integrations - Web content through controlled scraping operations **Embedding Management** - Private embedding models for sensitive content processing - Vector storage optimization within Postgres infrastructure - Semantic similarity algorithms for retrieval accuracy - Hybrid search combining vector and keyword approaches Retrieval Systems ^^^^^^^^^^^^^^^^^ :ref:`Retrievers Reference Manual ` components orchestrate search operations across diverse data sources, providing unified interfaces for knowledge discovery while maintaining source attribution and access controls. **Retrieval Strategies** - Vector similarity search for semantic content matching - Keyword-based search for exact term identification - Hybrid approaches combining multiple retrieval methods - Contextual filtering based on user permissions and data classification Assistant Framework ^^^^^^^^^^^^^^^^^^^ Intelligent :ref:`Assistants Reference Manual ` combine knowledge retrieval with language model capabilities to provide conversational interfaces for organizational knowledge. The framework supports complex reasoning workflows while maintaining audit trails for all interactions (see :ref:`Threads Reference Manual ` ). **Assistant Capabilities** - Multi-turn conversational interfaces with context preservation - :ref:`Tools Reference Manual ` integration for external system interactions - Reasoning workflows for complex query resolution - Audit logging for compliance and operational transparency Implementation Patterns ----------------------- Enterprise Knowledge Systems ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Organizations implement comprehensive knowledge management systems that unify structured and unstructured information sources through semantic search capabilities. **Typical Architecture** 1. **Data Integration Layer**: Connects multiple organizational :ref:`Configure Data Sources in Gen AI Builder ` including databases, document repositories, and knowledge management systems 2. **Processing Pipeline**: Transforms content into searchable formats with appropriate metadata and access control annotations 3. **Retrieval Interface**: Provides unified search across all integrated content with relevance ranking and source attribution 4. **Application Layer**: Delivers knowledge access through conversational interfaces and API integrations **Implementation Considerations** - Data governance policies for content classification and access control - Performance optimization for large document collections - Integration with existing authentication and authorization systems - Compliance requirements for audit trails and data lineage Retrieval-Augmented Generation Applications ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Applications combine retrieval capabilities with language model inference to provide context-aware responses based on organizational knowledge. **Development Workflow** 1. **Knowledge Base Construction**: Establish comprehensive document indexing with semantic embeddings (see :ref:`Create a Knowledge Base in Gen AI Builder ` and :ref:`Manage Knowledge Bases in Gen AI Builder ` ) 2. **Retrieval Configuration**: Implement search strategies optimized for application requirements 3. **Generation Integration**: Connect language models for response generation based on retrieved context 4. **Response Orchestration**: Coordinate retrieval and generation workflows with appropriate caching and optimization **Technical Requirements** - Low-latency retrieval for interactive applications - Context window management for large document processing - Response quality evaluation and improvement workflows - Integration testing across knowledge base updates Conversational AI Platforms ^^^^^^^^^^^^^^^^^^^^^^^^^^^ Interactive applications provide natural language interfaces for complex organizational workflows and information discovery. **Platform Components** - Multi-modal input processing for text, voice, and document interactions - Context management for extended conversational sessions - Tool integration for external system operations - Response personalization based on user roles and preferences **Operational Features** - Session management with conversation state persistence - Performance monitoring for response quality and latency - Usage analytics for application optimization and capacity planning - Security controls for sensitive information handling Development Framework --------------------- Knowledge Base Development ^^^^^^^^^^^^^^^^^^^^^^^^^^ Knowledge base creation involves data source integration, content processing, and semantic indexing to support efficient retrieval operations. **Development Process** 1. **Source Assessment**: Evaluate :ref:`Configure Data Sources in Gen AI Builder ` for content quality, update frequency, and access requirements 2. **Processing Pipeline Design**: Configure content transformation workflows including cleaning, chunking, and metadata extraction (see :ref:`Configure the Data Lake in Gen AI Builder ` ) 3. **Indexing Strategy**: Implement embedding generation and vector storage optimization 4. **Retrieval Testing**: Validate search quality through systematic evaluation procedures **Quality Assurance** - Content accuracy verification through systematic sampling - Retrieval relevance evaluation using standard metrics - Performance benchmarking under expected load conditions - Security testing for access control and data isolation Assistant Configuration ^^^^^^^^^^^^^^^^^^^^^^^ Assistant development requires careful orchestration of retrieval, reasoning, and response generation capabilities to deliver reliable user experiences (see :ref:`Create an Assistant in Gen AI Builder ` ). **Configuration Areas** - **Behavior Definition**: Establish assistant personality, capabilities, and operational boundaries - **Knowledge Integration**: Configure retrieval strategies and knowledge source prioritization (see :ref:`Retrievers Reference Manual ` ) - **Tool Integration**: Connect external systems and APIs for expanded functionality (see :ref:`Tools Reference Manual ` ) - **Response Optimization**: Tune generation parameters for appropriate response characteristics **Testing Strategies** - Conversational flow validation through systematic dialog testing - Knowledge accuracy verification across diverse query types - Integration testing for external tool interactions - Load testing for concurrent user scenarios API Development ^^^^^^^^^^^^^^^ Applications expose functionality through standardized APIs that support diverse client applications while maintaining consistent security and operational characteristics (see :ref:`Where to find SDK references and examples? ` ). **API Design Principles** - RESTful interfaces with clear resource modeling - Authentication and authorization integration with organizational systems - Rate limiting and quota management for resource protection - Comprehensive documentation and SDK support Integration Architecture ------------------------ EDB Postgres AI Ecosystem ^^^^^^^^^^^^^^^^^^^^^^^^^ Gen AI Builder integrates seamlessly with other EDB Postgres AI components to provide comprehensive AI application development capabilities. **Component Interactions** - \*\* :ref:`AI Factory Models ` \*\*: Governed access to language models and embedding models - \*\* :ref:`AI Accelerator Pipelines ` \*\*: Automated content processing and knowledge base updates - \*\* :ref:`Vector Engine (PGvector) ` \*\*: High‑performance semantic search within Postgres infrastructure - \*\* :ref:`EDB’s vision for AI with Postgres and Hybrid Manager ` \*\*: Orchestration and resource management External System Integration ^^^^^^^^^^^^^^^^^^^^^^^^^^^ Applications connect with existing organizational systems through standardized interfaces while maintaining security boundaries and audit capabilities. **Integration Patterns** - API-based connections for real-time data access - Batch processing for large data source synchronization - Event-driven updates for dynamic content management - Secure authentication delegation for user access control Operational Considerations -------------------------- Performance Characteristics ^^^^^^^^^^^^^^^^^^^^^^^^^^^ Gen AI Builder applications require careful performance optimization to deliver responsive user experiences while managing computational resources efficiently. **Performance Factors** - Knowledge base size affects retrieval latency and resource requirements - Model complexity influences response generation time and GPU utilization - Concurrent user load impacts system scalability and resource allocation - Network connectivity affects distributed component communication Security Framework ^^^^^^^^^^^^^^^^^^ Comprehensive security controls protect sensitive organizational knowledge while enabling appropriate access for authorized users and applications. **Security Components** - Multi-tenant isolation for different organizational units - Role-based access control aligned with existing authorization systems - Data encryption for knowledge base content and communication channels - Audit logging for all knowledge access and assistant interactions Governance Integration ^^^^^^^^^^^^^^^^^^^^^^ Built-in governance capabilities ensure applications comply with organizational policies and regulatory requirements throughout their operational lifecycle. **Governance Features** - Content approval workflows for knowledge base updates - Model usage tracking and compliance reporting - Data lineage maintenance for audit and troubleshooting - Performance monitoring aligned with service level objectives Getting Started --------------- Prerequisites ^^^^^^^^^^^^^ Gen AI Builder requires foundational infrastructure components including Hybrid Manager installation and appropriate compute resources for AI workloads. **Infrastructure Requirements** - Kubernetes cluster with sufficient compute and storage capacity - GPU resources for model serving and embedding generation - Network connectivity to organizational data sources - Integration with identity and access management systems Initial Implementation ^^^^^^^^^^^^^^^^^^^^^^ Organizations should begin with focused use cases that demonstrate value while establishing operational procedures for broader deployment. **Implementation Sequence** 1. **Pilot Knowledge Base**: Create initial knowledge base with well-defined content scope 2. **Basic Retrieval**: Implement simple search functionality with performance validation 3. **Assistant Prototype**: Develop conversational interface for pilot knowledge domain 4. **Production Deployment**: Scale successful prototypes with comprehensive operational controls Learning Resources ^^^^^^^^^^^^^^^^^^ **Conceptual Foundation** - :ref:`Gen AI Builder Hub ` - Platform overview and initial setup procedures - :ref:`Gen AI How-To Guides ` - Practical implementation guidance for common scenarios **Component Documentation** - :ref:`LibrariesΩ Manual ` - Comprehensive knowledge base development guidance - :ref:`Assistants Reference Manual ` - Intelligent assistant creation and configuration - :ref:`API Reference ` - Detailed technical documentation for development teams **Advanced Topics** - :ref:`Data Lake Reference Manual ` - Large-scale data processing and management - :ref:`SDK Documentation ` - Programmatic interfaces for custom development -------------- Gen AI Builder enables organizations to develop sophisticated AI applications while maintaining complete control over data, models, and inference operations within their existing infrastructure environments. .. toctree:: :maxdepth: 3 Usage--gen-ai--Getting_Started--index Usage--gen-ai--Manual--index Usage--gen-ai--Explainers--index Usage--gen-ai--How_to_prepare_data_and_kb--index Usage--gen-ai--How_to_bulid_assistants--index Usage--gen-ai--How_to_integrate_with_your_apps--index Usage--gen-ai--How_to_operate_and_observe--index