Infrastructure Scaling Plan

Roadmap to scale ForaCity to every city, county, and state in the United States.

Target Scale: United States

States

3,143

Counties

19,502

Cities

519K+

Elected Officials

22.7K

Total Jurisdictions

Phased Rollout

Key Metrics

Jurisdictions

~25

Subdomains

100+

Est. officers

5,000+

Est. contributions

Tech Stack

Multi-tenant DB schemaSubdomain routing per cityShared compliance engineBatch NetFile sync

Data Strategy

Automated ingestion pipelines: NetFile (all SD County filers), CivicPlus/Granicus (agendas), ArcGIS (all SD districts). Jurisdiction-scoped data isolation with shared analytics.

Deliverables

Multi-tenant database: jurisdiction_id on all tables, row-level filtering

Automated subdomain provisioning: {cityname}.foracity.com + {cityname}gov.foracity.com

Shared compliance engine with jurisdiction-specific thresholds

Batch NetFile sync across all SD County filers (FPPC jurisdiction)

County Board of Supervisors integration with district boundaries

Cross-jurisdiction analytics: compare compliance scores, participation rates

Onboarding workflow: city clerk uploads data → auto-provision subdomain

SD County pilot cities: Encinitas, Oceanside, Carlsbad, Vista, San Marcos

Data Ingestion Layer

Layer 1 of 5

NetFile Adapter (CA local campaign finance)

CalAccess Adapter (CA state campaign finance)

FEC Adapter (federal campaign finance)

CivicPlus/Granicus Adapter (agendas & minutes)

ArcGIS Adapter (district boundaries)

State-specific adapters (TEC, BOE, DOS, etc.)

PDF/OCR pipeline (Form 700, meeting minutes)

Manual upload portal (city clerk onboarding)

Processing & Compliance Layer

Layer 2 of 5

Pluggable Compliance Rules Engine

Conflict Detection Service

Contribution Threshold Monitor

Entity Resolution (match contributors across sources)

ML Anomaly Detection

Real-time Event Streaming (Kafka/NATS)

Storage Layer

Layer 3 of 5

TiDB Serverless (horizontally scalable MySQL)

Regional sharding (West/Central/East)

S3 for documents, images, media

Redis for caching and rate limiting

Elasticsearch for full-text search

Application Layer

Layer 4 of 5

tRPC API gateway (type-safe, auto-batching)

Subdomain router (auto-provision per jurisdiction)

Auth service (Manus OAuth + social logins)

Notification service (push, email, SMS)

LLM service (summaries, analysis, chat)

Presentation Layer

Layer 5 of 5

React 19 SPA with SSR capability

Mobile app (React Native, shared backend)

Embeddable widgets (for city websites)

Public API (REST + GraphQL)

Data export (CSV, JSON, PDF reports)

Cost Projections

Phase	Infra / mo	Data / mo	Team	Revenue Model
1. Escondido	$50	$0	1 dev	Grant / donation funded
2. SD County	$200	$100	2-3 devs	City subscriptions ($500/mo/city)
3. California	$2K	$1K	5-8 devs	Tiered SaaS ($200-2K/mo/jurisdiction)
4. Multi-State	$10K	$5K	15-20	Enterprise + API licensing
5. National	$50K	$20K	30-50	Platform fees + data API + grants

Key Architecture Decisions

Multi-Tenant vs. Multi-Instance

Multi-tenant with jurisdiction_id

Shared infrastructure reduces cost per jurisdiction. Row-level security via Drizzle query filters. Shard by region at Phase 3+.

Monorepo vs. Microservices

Monorepo with modular routers

tRPC routers act as service boundaries. Split to microservices only when team size exceeds 15. Current architecture supports 10K+ jurisdictions.

Real-time vs. Batch Processing

Batch with real-time overlay

Campaign finance data updates daily (NetFile/CalAccess). Real-time only for user-generated content (feed, Fora). Reduces infrastructure cost 80%.

Compliance Engine: Hardcoded vs. Rules Engine

Pluggable rules engine at Phase 4

Current CA Levine Act logic is hardcoded (simple, fast). At multi-state, extract to configurable rules: threshold amounts, lookback periods, entity types per state law.