Sava — Internal Documentation¶
Sava is a chartered trust company that builds its own technology stack — real-time dashboards, AI-powered workflows, and API-first architecture. We're building the infrastructure that makes trust administration feel like modern fintech.
This documentation is organized using the Diátaxis framework into four categories: explanation, how-to guides, reference, and tutorials.
Quick Start¶
- What Sava is: concept.md
- How the platform works: platform-architecture.md ← start here for the platform model
- Current build plan: Command Center Roadmap
- Tech stack & project structure: architecture.md
- Trust domain primer: trust-administration.md
- Terminology lookup: glossary.md
Documentation Map¶
Explanation¶
Understanding-oriented discussion of concepts and architecture.
| Document | What It Covers |
|---|---|
| ai-system.md | The Brain: 6 agents, infrastructure, evaluation, safety |
| api-architecture.md | Go API: Chi + Huma + sqlc, request lifecycle, type generation pipeline |
| architecture.md | Monorepo structure, tech stack, key patterns, dev workflow |
| concept.md | Product vision, problem/solution, core beliefs, competitive positioning |
| domain/directed-trusts.md | Directed trust structure, direction letters, fees, jurisdiction table |
| domain/trust-administration.md | Trust types, parties, P/I accounting, compliance, distribution standards |
| platform-architecture.md | The platform stack: Command Center + Sava Portal + Sava API + Trust OS |
| schema-management.md | Database migrations (goose), sqlc codegen, schema snapshots |
| orchestration-engine.md | HITL workflow engine: Go + asynq + Postgres FSM, outbox, approval gates |
| trust-instrument-storage.md | Trust instrument storage: three-layer model, extraction pipeline, knowledge graph |
How-To Guides¶
Task-oriented instructions for solving specific problems.
| Document | What It Covers |
|---|---|
| beta-program.md | Advisor preview: recruitment, testing, feedback, success metrics |
| dev-with-prod-secrets.md | Pull a single prod secret into a dev run (e.g. eval the prod Claude key) |
| dogfood-testing.md | Internal dogfood: team logs in as external personas and drives write flows on production |
| oncall-db-access.md | Connect to prod Postgres on-call without holding a long-lived DATABASE_URL |
| restore-from-backup.md | Restore an hourly Postgres dump from S3 into a target database |
| setup-aws-access.md | Set up AWS Identity Center access on a fresh laptop |
| run-henderson-demo.md | Step-by-step walkthrough: Henderson demo (attorney onboarding + grantor portal) |
| workflows.md | End-to-end flows: onboarding, distributions, compliance, comms |
Reference¶
Information-oriented technical descriptions and specifications.
| Document | What It Covers |
|---|---|
| data-model.md | All entities, relationships, enums, schema evolution |
| glossary.md | Trust terminology A–Z + platform terms |
| integrations.md | Third-party services: Supabase, Claude, S3, Postmark, DocuSign |
| personas.md | 7 user personas with needs, screens, and permissions |
| services.md | 11 Trust OS service modules, implementation status |
Roadmaps¶
Planning documents for each platform layer.
| Document | What It Covers |
|---|---|
| command-center-enhancements.md | Enhancement workstreams and UI component specs |
| command-center.md | 5-phase build plan for the trust officer operations interface (current focus) |
| sava-api-roadmap.md | Developer platform, API design, SDKs, webhooks |
| sava-portal-roadmap.md | External platform for attorneys, advisors, beneficiaries, CPAs |
| trust-detail-enhancements.md | Trust detail page tab specs and implementation status |
| trust-os-roadmap.md | Trust OS: services, accounting, AI agents, compliance |
| trust-relationship-mapping.md | Trust relationship and family grouping data model |
| user-model-and-roles.md | User model, roles, and permission matrix |
Tutorials¶
Learning-oriented guides that walk through a complete project or concept step by step.
| Document | What It Covers |
|---|---|
| getting-started.md | New developer setup: tooling, local stack, first PR |
Conventions¶
- Heading hierarchy:
#title,##sections,###subsections - Cross-references: Markdown links to other docs, e.g.,
(see [AI System](./explanation/ai-system.md)) - Tables for structured data, ASCII art for diagrams
- Tone: Direct, practical — written for an engineer shipping code
- Terms: Use definitions from glossary.md. If a term is missing, add it there.
- No personal content: These docs cover product, domain, and technical architecture only
- Layer awareness: When writing about features, be explicit about which layer (Command Center, Sava Portal, API, or Trust OS) owns the responsibility. See Platform Architecture.
- Organization: Follow the Diátaxis framework — tutorials, how-to guides, explanation, reference. Place new docs in the appropriate category.