Technology Stack

This document outlines the technologies used in the Stakeholder Agent System and the rationale behind each selection.

Overview

Layer	Technology	Purpose
Front-End	TypeScript, TanStack Start, Node.js	Browser-based chat interface
AI Service	Python, Pydantic AI, FastAPI	Prompt orchestration, AI guardrails
Data Store	Neon Postgres	Persistent storage
Communication	REST APIs	Service integration
Authentication	Neon Auth	Identity and session management
Hosting	Google Cloud Platform	Cloud infrastructure

Frontend

TypeScript

TypeScript provides static typing for the browser-based application, improving code reliability and developer experience through compile-time error detection and enhanced IDE support.

TanStack Start

TanStack Start serves as the foundation for the responsive chat interface. This framework was selected for:

• Modern React-based architecture with server-side rendering capabilities
• Type-safe routing and data fetching
• Optimized client-side state management for real-time chat interactions

Node.js

Node.js is used for the frontend build and development environment, enabling:

• Efficient package management with npm
• Fast development server with hot module replacement
• Compatibility with modern TypeScript tooling and libraries
• Seamless integration with TanStack Start and other frontend dependencies

Backend / AI Service

Python

Python serves as the primary language for the AI Service tier. The choice supports:

• Extensive AI/ML ecosystem compatibility
• Rapid development for prompt engineering workflows
• Strong async I/O support for non-blocking AI provider calls

Pydantic AI

Pydantic AI provides the framework for AI orchestration, offering:

• Structured prompt management and validation
• Type-safe data models for AI interactions
• Built-in support for context assembly and guardrails
• Token budget management and rate limiting capabilities

FastAPI

FastAPI is the web framework for the backend service, chosen for:

• High performance with async support
• Automatic OpenAPI documentation generation
• Easy integration with Pydantic for request validation
• Developer-friendly features that enhance productivity and maintainability
• Seamless injection of Neon Auth for authentication and session management

Data Storage

Neon Postgres

Neon is a serverless Postgres platform that provides durable persistence for all application state. This choice supports:

• ACID compliance for session data integrity
• Structured storage for user accounts, session metadata, and chat history
• Connection pooling for optimized performance
• Reliable state storage ensuring session continuity
• Serverless scaling that adapts to demand

For details on persisted entities and data flow, see Architecture.

Infrastructure

Google Cloud Platform

GCP provides the hosting infrastructure with:

• Scalable compute resources for the stateless backend design
• Network security and TLS encryption
• Health monitoring and logging capabilities

REST API Communication

The system uses REST APIs with JSON payloads for service integration:

• Clear contract definition between front-end and AI service
• Broad tooling support for testing and debugging
• Familiar patterns that support team productivity

Neon Auth

Neon Auth provides authentication and session management as part of the Neon platform, which:

• Reduces security implementation complexity
• Integrates seamlessly with Neon Postgres for unified data and identity management
• Simplifies compliance with security standards
• Handles token management and session validation

Architectural Alignment

The technology selections support the system’s primary quality attributes:

Quality Attribute	Technology Support
Availability	Stateless Python backend enables horizontal scaling; Neon Postgres provides durability with serverless scaling
Performance	Async I/O in Python; connection pooling; TanStack Start client optimization
Modifiability	Pydantic AI abstraction layer allows provider substitution; data-driven configuration
Security	Neon Auth for identity management; TLS throughout; Pydantic validation for input sanitization

---

Next Steps

Goal	Documentation
Understand system architecture	Architecture
Set up local development	Getting Started
Learn coding standards	Development Workflows