Add cloud hosting cost estimate documentation

[mriechers]

Summary

This PR adds comprehensive cost estimation documentation for running the PBS Wisconsin Editorial Assistant v3.0 in cloud infrastructure. The document provides detailed breakdowns of infrastructure costs, LLM API expenses, and deployment options to help stakeholders understand the financial implications of different hosting choices.

Key Changes

• New documentation file: planning/CLOUD_COST_ESTIMATE.md with 182 lines of detailed cost analysis
• Architecture overview: Documents the current system components (FastAPI backend, React dashboard, SQLite, background workers, MCP server)
• Cost comparison tables:
  • Infrastructure options (VPS, AWS EC2, Home VM) with monthly cost ranges
  • LLM API costs by model tier (Gemini Flash, Gemini Pro, Claude Sonnet)
  • Total monthly cost scenarios from light use ($17/mo) to production ($185/mo)
• Deployment options: Four detailed scenarios (Home VM, Cloud VPS, AWS EC2, Station Infrastructure) with cost breakdowns and maintenance requirements
• Scaling guidance: Thresholds and costs for when to add workers, migrate databases, or implement multi-region setups
• Production readiness checklist: 8-point checklist covering backups, monitoring, authentication, cost controls, and disaster recovery

Notable Details

• Emphasizes that LLM API costs dominate (75-90% of operational costs), making model selection the primary cost lever
• Documents built-in cost controls already configured in the system (DEFAULT_RUN_COST_CAP = $1.0, tiered model selection)
• Provides clear recommendations for minimal cost ($8-12/mo VPS) vs. production reliability (AWS EC2 with PostgreSQL)
• Includes scaling triggers and decision points for when infrastructure changes become necessary
• Targets both technical teams and stakeholders evaluating hosting options

This documentation supports informed decision-making about deployment strategy and helps set realistic expectations for operational costs.

https://claude.ai/code/session_01KX655cqiBjGnS9y4ScCYCf