Appearance

Technology Stack ​

Comprehensive overview of the technologies powering the Zeron platform.

TypeScript/Node.js Services ​

Conductor ​

  • TypeScript: Strongly typed orchestration logic
  • Redis 4.6: State management and job queues
  • Winston: Structured logging
  • Zod: Runtime type validation and schema definition
  • Axios: HTTP client for API calls
  • Claude Agent SDK: @anthropic-ai/claude-agent-sdk for headless agent spawning

Feedback Service ​

  • TypeScript: Type-safe application code
  • Node.js >= 18.0.0: Runtime environment
  • Prisma: Database ORM and migration tool
  • PostgreSQL: Primary database with pgvector extension
  • pgvector: Vector embeddings for semantic search

Learning Service ​

  • Node.js: JavaScript runtime

Public Website ​

  • Next.js 16: React framework with App Router
  • React 19: UI component library
  • Tailwind CSS v4: Utility-first CSS framework
  • TypeScript: Type-safe React components

Python Services ​

Dashboard ​

  • Python >= 3.8: Programming language
  • Streamlit: Interactive web application framework
  • Plotly: Interactive visualization library
  • Pandas: Data manipulation and analysis
  • NumPy: Numerical computing
  • NetworkX: Graph/network analysis

Discovery Engine ​

  • Python: Core language
  • BeautifulSoup4: HTML parsing
  • Playwright: Browser automation
  • PRAW: Reddit API wrapper
  • feedparser: RSS/Atom feed parsing
  • lxml: XML/HTML processing
  • readability-lxml: Content extraction

Databases ​

Neo4j 5.16 ​

  • Type: Graph database
  • Use Cases: Relationships, knowledge graphs, entity management
  • Deployment: Docker containerized (local dev) or Neo4j Aura (production)
  • Local Dev Credentials: neo4j/your-local-password (set in docker-compose.yml)
  • Ports: 7474 (HTTP), 7687 (Bolt)

PostgreSQL ​

  • Type: Relational database
  • Extensions: pgvector for embeddings
  • Use Cases: Feedback service data, semantic search
  • ORM: Prisma

Redis ​

  • Version: 4.6+ (6.0+ recommended)
  • Type: In-memory data store
  • Use Cases: Caching, state management, job queues
  • Persistence: AOF + RDB backups
  • Critical: Conductor state storage

AI/LLM Integration ​

Anthropic Claude ​

  • Sonnet 4.5: Primary model for high-quality tasks
    • Pricing: $3 input / $15 output per 1M tokens
    • Cached: $0.30 per 1M tokens
  • Haiku: Cost-optimized model for validation/testing
    • Pricing: $0.25 input / $1.25 output per 1M tokens
    • Cached: $0.03 per 1M tokens

OpenAI ​

  • GPT Models: Alternative LLM provider
  • Integration: Via LLM Router Bridge

PromptFoo ​

  • Purpose: AI conversation testing and optimization
  • Use Cases: Feedback service conversation quality validation
  • Features: Automated evaluation, regression testing

Claude Agent SDK ​

  • Package: @anthropic-ai/claude-agent-sdk@0.1.15
  • Purpose: Headless Claude Code agent spawning
  • Use Cases: Conductor builder/validator agents

Frontend Technologies ​

Next.js ​

  • Version: 16.x
  • Router: App Router
  • Features: Server components, streaming, optimized bundling

React ​

  • Version: 19.x
  • Use Cases: Feedback service client, public website
  • State Management: React Context, hooks

Tailwind CSS ​

  • Version: v4
  • Approach: Utility-first CSS
  • Features: Custom color schemes, responsive design

Infrastructure ​

Docker Compose ​

  • Services: Neo4j containerization
  • Configuration: docker-compose.yml
  • Management: docker-compose up -d

Vercel ​

  • Use Cases: Web service deployments
  • Features: Automatic deployments, serverless functions
  • Configuration: vercel.json files

Prisma ​

  • Version: 6.16.0
  • Features: Type-safe database client, migrations, studio
  • Schema: prisma/schema.prisma

Development Tools ​

Testing ​

  • Python: pytest, pytest-xdist (parallel), pytest-cov (coverage)
  • TypeScript: Jest, integration testing frameworks
  • PromptFoo: AI conversation testing

Code Quality ​

  • TypeScript: ESLint, Prettier
  • Python: Black (formatting), Flake8 (linting)
  • Pre-commit: Automated checks

Version Control ​

  • Git: Source control
  • GitHub: Repository hosting, PR workflows
  • GitHub Actions: CI/CD (referenced in conductor)

Deployment Architecture ​

Production Services ​

  • Feedback Service: Vercel deployment
  • Widget Demo: Live demonstration environment (preview.orderday.co)
  • Conductor: Self-hosted orchestration
  • Dashboard: Streamlit deployment

Infrastructure Services ​

  • Neo4j: Docker container
  • PostgreSQL: Managed database
  • Redis: Self-hosted/managed

Package Management ​

Node.js ​

  • npm >= 8.0.0: Package manager
  • Workspaces: Monorepo package management
  • Lock File: package-lock.json

Python ​

  • pip: Package installer
  • Requirements: requirements.txt
  • Virtual Environments: Recommended

Version Requirements ​

ToolMinimum VersionManaged By
Node.js18.0.0asdf (.tool-versions)
Python3.8System/pyenv
npm8.0.0Comes with Node.js
DockerLatestSystem
Redis4.6Homebrew/system
PostgreSQLLatestSystem/managed

Cost Optimization ​

The platform implements strategic cost optimization:

  • Model Selection: Haiku for validation (~90% cheaper than Sonnet)
  • Prompt Caching: 90% cost reduction on cached tokens
  • Budget Tracking: Per-ticket cost monitoring in Conductor
  • Token Optimization: Minimize unnecessary AI calls

Typical Costs (per ticket):

  • Simple: $0.50 - $2.00
  • Complex: $2.00 - $5.00
  • Edge cases: $5.00 - $10.00

Part of the Zeron Platform | Built with VitePress

Copyright © 2025 Zeron