Agent Factory Deployment

Service Overview

The EDB Agent Factory Deployment engagement delivers an installed, configured, and validated EDB Agent Factory environment. This includes the Agentic Workflow Engine (LangFlow), Model Serving infrastructure, Vector Engine (including pgvector, hybrid search and AI pipelines), MCP Server, and Knowledge Base infrastructure — the complete foundation required to build and deploy production AI applications on EDB Postgres AI.

Deployment engagements focus exclusively on installation, configuration, and technical validation. When the engagement concludes, your environment is operational and ready for AI application development or a follow-on Agent Factory Enablement engagement.

Scope of Service

Onboard

• Kickoff meeting to confirm scope, validate infrastructure prerequisites, and establish communication protocols
• Review compute infrastructure: GPU availability, CPU specifications, storage, network topology
• Validate EDB Postgres AI licensing and HM or VM connectivity

Installation

• Deploy Agent Factory components via Hybrid Manager or stand alone(Agentic Workflow Engine, Model Serving, Vector Engine)
• Install and configure MCP Server endpoints
• Create pgvector + AIDB extension across target Postgres databases
• For Medium+: configure GPU-based model serving infrastructure with validated NVIDIA NIMs or HuggingFace models
• For Large+: install and configure Bring Your Own Model (BYOM) capabilities

Configuration

• Configure model catalog with pre-validated models appropriate for deployment tier
• Set up Hybrid Search infrastructure 
• Configure Knowledge Base data pipeline infrastructure for connected data sources
• Establish MCP Server tool registry and endpoint security
• For Large/X-Large: configure high availability and failover settings

Validation

• Execute end-to-end smoke tests across all installed components
• Validate model serving response (latency, throughput benchmarks)
• Confirm pgvector indexing and query performance
• Test MCP Server connectivity and tool discovery
• Validate Knowledge Base pipeline with sample ingestion
• Review and document Standard Operating Procedures for component management

Project Closure

• Formal review of completion status against agreed scope
• Deliver configuration documentation and handoff to customer team

Deliverables

• Fully installed and operational EDB Agent Factory environment
• Configured model serving with validated model catalog
• Operational MCP Server with documented endpoints
• pgvector-enabled Postgres database(s) with validated indexing
• Knowledge Base pipeline infrastructure (ready for data ingestion)
• Configuration documentation and operational SOPs

Roles and Responsibilities

EDB Project Manager: responsible for engagement planning, prerequisite coordination, schedule management, and project closeout.

EDB Solution Architect: reviews infrastructure requirements, designs deployment architecture, and validates installation against EDB Agent Factory best practices.

EDB Senior Consultant: technical lead for all installation, configuration, validation, and documentation activities.

Customer Team: provides infrastructure access and participates in validation activities. Key roles:

• Platform / Infrastructure Engineer
• Database Administrator
• Security / Network Administrator
• IT Operations Lead

Assumptions

• A kickoff call will confirm scope and schedule prior to work beginning.
• This engagement is delivered remotely unless otherwise agreed.
• EDB Hybrid Manager is deployed and operational in the customer environment prior to engagement start.
• Compute infrastructure meeting Agent Factory hardware requirements is available and accessible.
• For GPU-based tiers: customer has provisioned GPU-enabled nodes meeting NVIDIA requirements.
• EDB Postgres AI software licenses are in place.
• This engagement deploys Agent Factory infrastructure only; AI use case development is out of scope (see Agent Factory Enablement).
• Customer network and security policies permit required container image downloads, or customer provides a private registry.
• Customer will provide timely feedback on deliverables. Items without commentary within 5 business days are deemed accepted.
• Customer will not provide Personal Data as defined by applicable law.

Prerequisites

• EDB Hybrid Manager (version 1.3 or later) is deployed and accessible.
• Kubernetes cluster with sufficient capacity is available (EDB-supported distribution).
• For GPU model serving: NVIDIA GPU nodes with appropriate drivers (CUDA 12.x+) are provisioned.
• Customer has cluster-admin access to the Kubernetes environment.
• Container registry access is available (public or private).
• pgvector-compatible EDB Postgres version is deployed.
• Network connectivity between Agent Factory components and target Postgres instances is confirmed.
• Customer's security team has reviewed and approved component deployment.

Validation Test Cases